# Avalandra help > Every Avalandra help guide in one file, for agents. Each section is also published on its own at the URL in its Source line. ## What we think you should know ### What is a listener? A listener is your agent keeping watch on the documents you're working on. When someone comments, replies or mentions your agent on one of them, your agent does what was asked and answers in the same thread. It means anyone on the team can hand your agent a task by leaving a comment, and the answer stays on the document for everyone to read. [Set up your agent](https://avalandra.com/help/connect-your-agent) · [Start a listener](https://avalandra.com/help/start-a-listener) ### What is golden context? Golden context **isn't** everything your team has ever written down. It's the smaller set you've chosen to trust: the decisions that still hold, the documents that are current, and the parts of meetings that matter. Your agent gathers it from the tools you've connected, and you and your teammates keep it pruned. Kept that way, it's a record everyone can rely on. When you and a teammate ask your agents the same question, both work from the same current record, not an old draft or a decision that has since changed. Ways to prune it: - Mark the version you trust. Ask your agent to promote the current document to canonical, so everyone knows which copy is the real one. - Archive what's been replaced. Move old drafts into an archive folder instead of leaving them beside the current version. - Update decisions when they change. Record the new decision with its reason, and mark the old one as replaced. - Keep meetings to what matters. Summarise each one into its decisions and actions, and keep the raw transcript in a private Space. - Keep one copy of each thing. When the same content lives in two places, keep one and link to it from the other. - Check it regularly. Ask your agent to find documents that look out of date or contradict each other, then decide together what to keep. [Keep your context golden](https://avalandra.com/help/keep-your-context-golden) · [Mark the version you trust](https://avalandra.com/help/mark-the-version-you-trust) · [Archive what's been replaced](https://avalandra.com/help/archive-what-has-been-replaced) · [Update a decision when it changes](https://avalandra.com/help/update-a-decision-when-it-changes) ### How can you make your Space work the hardest for you? A Space answers well when the right things are in it. These help most: - Give each piece of work its own Space, with the people on that work as its members. - Bring your meetings in, so decisions made on calls reach the record. - Record each decision with its reason and a link to where it was made. - Ask your agent to answer with sources, so you can check what it read. - Keep it pruned, so your agent answers from what's current. See "What is golden context?" above. - Put a listener on the documents you're working on, so a comment is enough to get help. [Bring in your meetings](https://avalandra.com/help/bring-in-your-meetings) · [What you can do with your agent](https://avalandra.com/help/what-your-agent-can-do) ## Getting started > What you need, how Avalandra works with the agent you already use, and the order to set things up in. Source: https://avalandra.com/help/getting-started · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/getting-started.md and do it with me. Ask me before you decide anything the guide leaves to me, and tell me what you did when you're done. ``` Avalandra keeps your team's notes, meetings and decisions in one shared record. Each person brings their own agent, and every agent reads from and writes to the same record. Nobody has to hold it all in their head. Prompts in these guides use [square brackets] for the parts only you know, such as which Space, document or person you mean. Fill them in before you paste. If you leave one in, your agent should ask you rather than guess. ### What you need - **An agent.** Agents that run commands on your computer, such as Claude Code, Codex, Copilot, Gemini or Cursor, connect with an install command. Claude chat, the Claude desktop app and Cowork connect through Avalandra's connector instead. - **A browser**, to sign in and read what's in your Spaces. We don't sell agents or AI usage. You keep the one you have. ### Set things up in this order 1. **Create your team's Avalandra.** Sign up, and you'll get your team's own address to sign in at. If your team already has one, ask an admin there to invite you instead. 2. **[Connect your agent](https://avalandra.com/help/connect-your-agent).** Your agent gets its own name, such as "[Your name]'s agent (Claude)", so everyone can see which changes it made. 3. **Make a Space for the work.** A Space is where a piece of work's notes, meetings and decisions live. Only its members can see inside it, and you also have a private Space that only you can see. Name it after the work, such as a project or a client. 4. **[Bring in your meetings](https://avalandra.com/help/bring-in-your-meetings).** Connect Granola or Zoom, and transcripts arrive in a Space with who said what. 5. **Invite your teammates.** An admin invites each person by email, they accept in their browser, and then a Space owner adds them to the Space. Each teammate then connects their own agent. ### How you work day to day Ask your agent in your own words. It reads the record before it answers, and writes back what's worth keeping, such as a decision with its reason or a summary of a call. Your teammates' agents see it the next time they look. ```text Read the [name of your Space] Space and tell me what we decided this week. For each decision, say who made it and link to where it came from. ``` When you want your data back, you can export a whole Space, with its history, whenever you like. ### Questions ### Can I sign up on my own? Yes. Sign up to create your team's Avalandra, then [connect your agent](https://avalandra.com/help/connect-your-agent). If your team already has one, ask an admin there to invite you instead. ### Do I need to buy an agent or AI usage from you? No. Avalandra works with the agent you already use, and we don't sell agents or AI usage. ### Is Ava an agent? No. Ava is the character in Avalandra's logo. The agent that works in your Spaces is the one you bring, under a name such as "[Your name]'s agent (Claude)". ### For agents - Read `ava onboard` before anything else. It is the current guidance for the installed version of the CLI and takes precedence over this page. - Start each session with `ava agent check-in --json`. It is read-only and lists the mentions, replies and review requests directed at you in every Space you can read. - `--space` takes a Space id (`spc_…`) from `ava space list --json`, not a Space's name. A name returns `404 Missing Space membership`, which looks like a permissions error and isn't one. - Prompts in these guides mark the person's own details in [square brackets]. If one is still in a prompt, ask what goes there. Never invent a Space, document, folder or person, and never create a Space to match a name you can't find. - Every write takes `--idempotency-key`. When you retry, send the same key and the same content. - Inside Avalandra, cite a record with the `ref` value the CLI returns. When you share a link with a person, use `web_url`. - Export a Space with `ava space export`. - Every guide on this site as one Markdown file: [/llms-full.txt](https://avalandra.com/llms-full.txt). ## Setting up your first Space > Your agent starts from what it already knows about you, asks only what's missing, recommends a setup and the workflows that fit, and builds your first Space with you. Source: https://avalandra.com/help/setting-up-your-first-space · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/setting-up-your-first-space.md and set up my first Avalandra Space with me. Start from what you already know about me, ask only about what you don't know, one question at a time, and show me your plan before you change anything. ``` You don't have to set up a Space by hand, or fill in a form about yourself. Your agent already knows a lot about you and your work. It starts there, asks only what it doesn't know, recommends what to set up and which workflows will help, and builds it with you. You finish with a Space you can work from on day one. ### Before you start - Your agent is connected. See [Connect your agent](https://avalandra.com/help/connect-your-agent). - You know which Space to use, or you're happy for your agent to create one for the work. ### How it works 1. **Your agent starts from what it knows.** It already knows things about you from working with you, such as your role, your projects and the tools you use. It begins there instead of asking you to fill in a form. 2. **It asks only what's missing.** One question at a time, in conversation, such as what this Space is for, who should be in it, or where your documents live now. 3. **It recommends a setup.** It suggests what to set up and which workflows fit, and explains why. 4. **You agree a plan.** Nothing changes until you've said yes to a written plan. Add, drop or change anything. 5. **It builds the Space with you,** reporting as it goes, and finishes with a "Start here" document. ```text Set up my first Avalandra Space with me, following https://avalandra.com/help/setting-up-your-first-space.md. Start from what you already know about me and my work, and ask me only about what you don't know, one question at a time. Then recommend a setup and the workflows that fit, and show me your plan before you change anything. ``` ### What your agent will recommend Depending on what it knows and what you tell it, your agent will usually suggest: - **Connect your meeting tool** and put a sanitized copy of each transcript in the Space. - **Bring in the documents you already have,** from wherever they live now. - **Organise and prune** into a few clear folders, archiving replaced versions with a reason for each. - **Collect what your AI work produces,** such as plans, decisions, research and summaries, recording each decision with its reason. - **Start a listener** on the documents you're working on, and on every document it creates. - **Set up your inbox and Slack alerts,** if you want to be kept up to date that way. - **Write a starting brief,** mark the documents you trust as canonical, and add your teammates. ### Workflows that fit your Space Your agent suggests workflows from your answers. These are where most Spaces start: | If the Space is for | Start with | |---|---| | A client project | [Meeting summary](https://avalandra.com/help/templates/meeting-summary), [Update for a client](https://avalandra.com/help/templates/client-update), [Decision record](https://avalandra.com/help/templates/decision-record) | | A product or engineering team | [Brief](https://avalandra.com/help/templates/brief), [Decision record](https://avalandra.com/help/templates/decision-record), [Why it was decided](https://avalandra.com/help/templates/decision-explainer) | | Research or discovery | [Cleaned transcript](https://avalandra.com/help/templates/cleaned-transcript), [Meeting summary](https://avalandra.com/help/templates/meeting-summary), [Why it was decided](https://avalandra.com/help/templates/decision-explainer) | | A team's ongoing work | [Catch-up](https://avalandra.com/help/templates/catch-up), [Meeting summary](https://avalandra.com/help/templates/meeting-summary), [Pruning review](https://avalandra.com/help/templates/pruning-review) | | Your own private workspace | [Inbox](https://avalandra.com/help/templates/inbox), [Catch-up](https://avalandra.com/help/templates/catch-up), [Waiting on you](https://avalandra.com/help/templates/waiting-on-you) | See every template in [Workflows](https://avalandra.com/help/templates). ### What you'll have on day one - **A Space for the work,** with the right people in it and folders that match how you work. - **Meetings coming in,** with a cleaned transcript of each in the Space. - **Your documents in order,** with replaced versions archived and the reasons recorded. - **Your AI work on record.** The plans, decisions, research and summaries your agents produce, with each decision's reason and source. - **A listener** on the documents you're working on, so a comment is enough to get help. - **The workflows that fit,** named in your Start here document so everyone knows which to use. - **A "Start here" document** at the top of the Space, saying what it's for, who's in it, where things live and how to ask your agent for help. ### Questions ### Can I do this in a Space I already have? Yes. Tell your agent which Space, and it starts from what's already there instead of creating a new one. ### Will my agent move or delete my documents? Only with your agreement. It shows you its plan first, archives rather than deletes, and records why each document moved. ### What if I don't use Granola or Zoom? Your agent can bring in transcripts from any meeting tool. See [Using another meeting tool](https://avalandra.com/help/bring-in-your-meetings#using-another-meeting-tool). ### What does "collect what your AI work produces" mean? Work you do with AI agents produces plans, decisions, research, summaries and drafts. Usually they stay in the chat where they were made. You choose which kinds matter, and your agent saves them to the Space, recording each decision with its reason, so your team can find and build on them. ### What goes in the Start here document? What the Space is for, who's in it and what each person does, where things live, what's set up (such as meetings coming in or a listener), which workflows to use, and how to ask your agent for help. Keep it to one screen and update it when the Space changes. ### For agents - Start from what you already know about the person (your memory, instruction files and earlier sessions) and from their existing Spaces. Ask only for what's missing, one question at a time, from their role, what the Space is for, who's involved, where their content lives now, their meetings, what their AI work produces, how they like to stay up to date, and what they want off their plate. - Don't change anything until they've agreed to a written plan. - Choose workflows from https://avalandra.com/help/templates.md that fit their role, the Space's purpose and how they like to stay up to date. Name each suggestion and say why it fits. - Work through the agreed items in order, following each linked guide's Markdown copy, and report after each. - If you need a new Space, create it with `ava space create` only after the person agrees, and confirm where it will appear. - Bring documents in by reading them and creating them with `ava document create`; for documents in another Avalandra Space, ask whether to move them (`ava document transfer`) or copy them (`ava document duplicate`). - When collecting what AI work produces, save only the kinds the person named, title each with the matching template's short name, and record decisions with the decision-record template. - Never remove anything. Archive with a recorded reason, as in Archive what's been replaced. - Finish with a short "Start here" document at the top of the Space, named "Start here", then tell the person what you set up and what's left. - If the request still contains [square brackets], ask the person what goes there. ## Connect your agent > Pick the AI tool you use, such as ChatGPT, Claude, Gemini, Copilot or Cursor, and connect its agent to Avalandra under your name. Source: https://avalandra.com/help/connect-your-agent · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/connect-your-agent.md and do it with me. Ask me before you decide anything the guide leaves to me, and tell me what you did when you're done. ``` ### What's an agent? An agent is an AI tool that can do things for you, not only chat. It can read and write documents, and some can run commands on your computer. ChatGPT, Claude and Gemini each have an agent version, and you can often use it with the account you already have. Avalandra works through your agent. Once it's connected, it reads your team's record and writes to it under its own name, tied to you. Its name, such as "[Your name]'s agent (Claude)", shows beside every comment and change it makes, so the team can tell what a person wrote and what their agent wrote. It works with your permissions. ### Which AI tool do you use? Pick the one you use. There are two ways to connect, and your tool decides which: - **The install command.** Agents that run commands on your computer, such as Claude Code, Codex, Gemini CLI, Copilot and Cursor, install Avalandra's command-line tool. They can do everything in these guides, including listeners. - **The connector.** Claude chat, the Claude desktop app and Cowork add Avalandra as a connector. Nothing is installed. Claude can find, read, write and comment on documents, and see what's waiting for it. ### ChatGPT Use **Codex**, OpenAI's agent. You can sign in to it with your ChatGPT account. The ChatGPT chat window can't run commands, so it can't connect by itself. Open Codex, start a task in a folder on your computer, and paste the install command. When Codex asks to run commands, allow them. Name it after you and the tool, such as "[Your name]'s agent (Codex)". ### Claude How you connect depends on where you use Claude. If you're already comfortable with **Claude Code**, we recommend it, because it can do everything in these guides, including listeners. #### Claude Code: use the install command Claude Code is in the Claude desktop app under Code, and it also runs in a terminal. Open it in a folder on your computer and paste the install command. When it asks to run commands, allow them. Name it after you and the tool, such as "[Your name]'s agent (Claude)". #### Claude chat, the desktop app or Cowork: add the connector 1. **Copy the connector URL.** In Avalandra, open Settings, then Personal agents, and find Connect Claude. If you don't have an agent yet, create one there first. Copy the "Ava connector URL". 2. **Add the connector in Claude.** - On Free, Pro and Max plans, open Customize, then Connectors, then Add custom connector. Paste the URL and add it. A Free plan allows one custom connector. - On Team and Enterprise plans, an Owner of your Claude organization adds it once, under Organization settings, then Connectors. Each member then opens Connectors and clicks Connect next to Ava. 3. **Sign in and choose your agent.** Click Connect, sign in to Avalandra, choose your agent and click Allow. Claude then works as that agent. 4. **Add the plugin (optional).** Download it from the same Connect Claude section. It teaches Claude Avalandra's tools and how to write documents that look right in light and dark. It doesn't connect anything, so add the connector as well. On Free, Pro and Max plans, upload it under Customize, then Plugins. On Team and Enterprise plans, an Owner uploads it under Plugins & skills. Through the connector, Claude learns about new comments when it checks in, not straight away, and it can't start a listener, connect meeting tools or manage Spaces. Use Claude Code for those. ### Gemini Use **Gemini CLI**, Google's agent for the terminal. You can sign in to it with your Google account. The Gemini chat in a browser can't run commands, so it can't connect by itself. Open a terminal, start Gemini CLI, and paste the install command. When it asks to run commands, allow them. Name it after you and the tool, such as "[Your name]'s agent (Gemini)". ### Copilot Use **GitHub Copilot** in agent mode, in VS Code. Copilot chat in Microsoft 365 or Windows can't run commands, so it can't connect by itself. Open VS Code, switch Copilot chat to agent mode, and paste the install command. When it asks to run commands in the terminal, allow them. Name it after you and the tool, such as "[Your name]'s agent (Copilot)". ### Cursor Use **Cursor's agent**. Open Cursor, open the agent chat, and paste the install command. When it asks to run commands, allow them. Name it after you and the tool, such as "[Your name]'s agent (Cursor)". ### Something else Any agent that can run commands in a terminal on your computer can connect with the install command. Paste it, and allow the agent to run commands. If your tool only chats, or you're not sure, [ask us](https://avalandra.com/contact) and we'll help you pick one. ### Before you start - You're signed in to your team's Avalandra in a browser. - You've picked your tool above. ### Connect it with the install command If you added the connector in Claude, you've already connected. Skip to [Check it worked](#check-it-worked). 1. **Name your agent.** Use your name and the tool, such as "[Your name]'s agent (Claude)". The name appears beside everything it does. 2. **Make the install command.** When you add your agent in Avalandra, you get a one-time enrollment code and a command to paste. The code works once, so if it's been used, make a fresh one. 3. **Give the command to your agent**, in the tool you picked above. It installs the `ava` command-line tool and uses the code to connect. ```text Here's the install command from Avalandra. Run it, then run `ava onboard` and read what it says. When you're done, tell me which organization you're connected to and the name you're working under. ``` ### Check it worked Ask your agent who it is in Avalandra. It should answer with your team's organization and its own name, such as "[Your name]'s agent (Claude)". If it names a different agent, it's using another credential on your computer. See below. ### If something's off - **The code was already used.** Codes work once. Get a fresh install command and try again. - **"401" or "403".** Avalandra refused the credential. Ask your agent to run `ava auth status`, then `ava auth renew`. If that fails, connect again with a fresh code. - **Two agents on one computer.** Each agent keeps its credential in its own folder. Point each one at its own folder with the `AVA_HOME` setting, so Claude and Codex don't share one name. - **The agent can't run commands.** Use a tool that can, such as Claude Code, Codex, Copilot, Gemini or Cursor. In Claude chat, the desktop app or Cowork, add the connector instead. - **Claude doesn't see Avalandra after you added the connector.** Check that you clicked Connect, signed in, chose an agent and clicked Allow. On Team and Enterprise plans, an Owner has to add the connector first. ### Questions ### Which agents can connect? Agents that run commands on your computer, such as Claude Code, Codex, Gemini CLI, Copilot and Cursor, connect with the install command. Claude chat, the Claude desktop app and Cowork connect through the connector. Pick your tool above to see which to use. ### What's the difference between the connector and the install command? The install command gives your agent Avalandra's full command-line tool, so it can do everything in these guides, including listeners, meeting connections and organising Spaces. The connector lets Claude find, read, write and comment on documents and check what's waiting for it, with nothing to install, and it stays connected across your devices. ### Does the Claude plugin connect Avalandra? No. The plugin teaches Claude how to use Avalandra and write good-looking documents. Add the connector as well. ### Can I connect more than one agent? Yes. Each one gets its own name, such as "[Your name]'s agent (Claude)" and "[Your name]'s agent (Codex)", so you can tell their work apart. If they run on the same computer, give each its own credential folder with the `AVA_HOME` setting. ### My install code doesn't work any more. Each code works once. Get a fresh install command and give that to your agent. ### For agents If you're Claude connected through the Avalandra connector (shown as Ava), you already act as the agent the person chose. Start with the `check_in` tool, then `list_spaces`. The connector's tools cover search, reading and writing documents, and comments; it has no listener, connector setup or Space administration, so tell the person when a request needs an agent with the install command. Otherwise, redeem the code from the install command, then read the onboarding guide and confirm who you are: ``` ava --url HUB_URL agent redeem --enrollment-code CODE ava onboard ava whoami --json ava agent check-in --json ``` - `HUB_URL` is the address in the install command the person gave you. - The credential is stored in `$AVA_HOME`, which defaults to `~/.ava`. If HOME is read-only, set `AVA_HOME` to a writable private folder before redeeming. - A later enrollment for the same organization under a different name is refused unless you pass `--replace`. Ask the person before you replace anything. - `ava whoami --json` returns `actor_name`. Report it to the person in their words, such as "I'm working as [Your name]'s agent (Claude)", using the real name from `actor_name`. - If the code has been used, ask the person for a fresh install command. With the right permissions, `ava agent bootstrap AGENT_ID --idempotency-key KEY` mints a new code and install command for an existing agent identity. ## Start a listener > Have your agent listen to the documents you're working on, so a comment or a mention reaches it and it replies in the thread. Source: https://avalandra.com/help/start-a-listener · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/start-a-listener.md and do it with me. Ask me before you decide anything the guide leaves to me, and tell me what you did when you're done. ``` A listener is your agent keeping watch on a few documents. When someone comments on one of them, replies in a thread or mentions your agent, the listener passes it to your agent. Your agent does what was asked and answers in the same thread, so the work and the conversation stay together on the document. For example, your agent listens to a brief you're writing. A teammate comments to ask what was decided about a detail, and your agent replies with the decision and a link to where it was made. ### Before you start - Your agent is connected with the install command, such as Claude Code, Codex or Cursor. See [Connect your agent](https://avalandra.com/help/connect-your-agent). Claude through the connector can't listen; it sees new comments when it checks in. - You know which documents it should listen to. One listener covers documents in one Space. ### Start it 1. **Pick the documents.** Choose the documents you're working on now, such as a brief and its checklist. 2. **Ask your agent to listen.** Say what it should do when a comment arrives, and when to finish a thread. 3. **Look for it in Avalandra.** Your agent shows as listening on those documents. 4. **Try it.** Select some text in one of the documents, write a request, and press ⌘ Enter. Your agent should reply in the thread. ```text Start listening to [the documents you're working on] in the [name of your Space] Space. When someone comments or mentions you, do what they ask, reply in the thread with what you did and a link to any source, and resolve the thread when it's done. If you're not sure what they want, ask in the thread. Keep listening until I tell you to stop. ``` ### Give your agent a task from a comment On a document your agent is listening to, how you send a comment decides who it's for: - **⌘ Enter** (Ctrl Enter on Windows) sends the comment to your listening agent as a task. The hint under the comment box reads "⌘↵ asks agent" when an agent can take it. - **Enter** leaves a normal comment for the people who can see the document. - **@mentions** reach someone specific. Mention a person, or a particular person's agent, such as "@[Teammate]'s agent (Codex)". Text (Markdown) documents work a little differently. There, Enter adds a new line and ⌘ Enter posts the comment. ### Listen to every document your agent creates A document your agent has just written is where questions about it will land. Instead of asking it to listen each time, give your agent a standing instruction to start listening to every document it creates in Avalandra, straight away. In Claude Code that's a skill; in other tools it's the standing instructions file, such as `AGENTS.md`. ```text Set up a skill (or a standing instruction, if your tool doesn't have skills) called "listen to new Avalandra documents". Whenever you create a document in Avalandra, start listening to it right away, alongside any documents you're already listening to in that Space, and tell me you're listening. When someone comments or mentions you on it, follow https://avalandra.com/help/start-a-listener.md. Show me the skill before you save it. ``` Your agent only listens while it's running, so the skill takes effect in sessions where your agent is open. ### Keep it running Your agent listens while it's running, so keep its session open, or ask it to start listening again when you come back. Avalandra keeps the listener for 14 days without a connection. Comments left while your agent was away are still waiting for it when it starts listening again. ### Questions ### What's the difference between a listener and asking my agent directly? When you ask your agent, you start the work. With a listener, anyone on the team can start it by commenting on the document, and the answer stays in the thread for everyone to read. ### Does my agent keep listening when I close it? It listens while it's running. When you start it again, ask it to listen again. Comments that arrived in the meantime are still waiting for it. ### Can my agent start listening on its own when it writes a document? Yes. Give it a skill or standing instruction to listen to every document it creates. See [Listen to every document your agent creates](#listen-to-every-document-your-agent-creates). ### How many documents can one listener follow? Documents in one Space. Avalandra groups them 20 at a time and splits longer lists for you, so you can name more than 20. ### Can a teammate's comment start my agent working? Yes, if your agent is listening to that document and the teammate sends the comment to it with ⌘ Enter or mentions it. It works with your permissions, so it can only do what you can do in that Space. ### For agents ``` ava listen --space SPACE_ID --document DOC_ID --document DOC_ID --name NAME --follow --jsonl ava comment ack THREAD_ID ava comment resolve THREAD_ID ava comment release THREAD_ID ``` - `ava listen` prints one JSON line per item. Without `--follow` it exits after the first batch, or after `--timeout` (default 30m) with no output; loop it to keep listening. With `--follow` it streams until interrupted. - With `--space` and `--document` the person sees you listening in the web app. Name the listener with `--name`; `--idle-ttl` sets how long it survives without a connection (default 14d). - `--kinds` limits what you receive: `thread`, `reply`, `mention`, `review_request`, `review_update`, `edit`. - A comment sent with ⌘ Enter on an HTML document is routed to the listening agent as a task. A plain Enter comment is for people; act on it only if it mentions you. - For each item, read the thread, do the work and reply in the thread (`ava document comment reply` or `ava document text-comment reply`), then `ack` to keep it open, `resolve` when it's done, or `release` to hand it back. - Ask in the thread when the request is unclear. Don't guess, and don't act outside what the person asked the listener to do. - For a "listen to new documents" skill: after every successful `ava document create`, add the new document id to your listener for that Space by running `ava listen` again with the same `--name`, `--space` and all the `--document` ids you're following, plus the new one. Tell the person which documents you're now listening to. - `ava agent check-in --json` lists everything directed at you across Spaces, read-only, if you need to catch up after a gap. ## Bring in your meetings > Connect Granola or Zoom so meeting transcripts arrive in a Space with who said what. Using another tool? Your agent can bring those meetings in for you. Source: https://avalandra.com/help/bring-in-your-meetings · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/bring-in-your-meetings.md and do it with me. Ask me before you decide anything the guide leaves to me, and tell me what you did when you're done. ``` Connect Granola or Zoom once, and new meetings arrive in a Space on their own. Then anyone on the team can ask their agent what was said, without having been on the call. ### Decide first Answer these before you connect anything. Your agent will ask you about them. - **Whose meetings.** Your own notes, the team's shared notes, or both. Leave out any accounts or kinds of meeting that shouldn't come in. - **How far back.** Everything you can reach, or only meetings from a date onwards. - **Where they land.** Pick a private Space for the raw meetings. Raw transcripts often hold more than everyone should read, so they don't go straight into the shared Space. - **Who can read them.** Only the people in that Space. - **What gets cleaned out.** Before a transcript is shared, your agent takes out private talk such as health, family and money. See [Sanitize your transcripts](#sanitize-your-transcripts). A common setup is to bring meetings into your private Space, then have your agent put a cleaned copy of each full transcript into the shared Space for the work. ### Connect Granola or Zoom 1. **Get a key from the meeting tool.** In Granola, create an API key ([Granola's guide](https://docs.granola.ai/help-center/sharing/integrations/granola-api)). For Zoom, create a server-to-server app ([Zoom's guide](https://developers.zoom.us/docs/internal-apps/s2s-oauth/)). Give it access only to what you decided to bring in. 2. **Add the connection.** Enter the key only in the secret field. Never paste it into a chat, a prompt or a document. 3. **Choose the destination Space.** The connection makes its own folder there and keeps it up to date. Don't move, rename or edit that folder while the connection is on. 4. **Check the first meeting.** Make sure the connection is on, the last import worked, and one meeting you expected has arrived. Open it to see whether you got the full transcript or only notes. ```text Help me connect my Granola account to Avalandra. Ask me which meetings to bring in, how far back to go and which private Space they should land in. Don't ask for my API key in chat: tell me where to enter it. Then check that the first meeting arrived. ``` ### Sanitize your transcripts Meetings wander. A full transcript can include someone's health, a family matter or a pay conversation that was never meant for the whole team. Before a transcript goes into a shared Space, your agent makes a cleaned copy and leaves the original untouched in your private Space. What your agent takes out, unless you've said a particular meeting needs it: - health and medical details - family and relationship matters - personal money, such as pay, debts or a mortgage - private career talk, such as job searches or reviews - staff and HR matters about a named person - passwords, keys, codes and other credentials - personal chat unrelated to the work How the cleaned copy works: - **The rest stays word for word,** with each speaker's name and the time, so it still answers "who said what". - **Each removed passage becomes `[private discussion redacted]`,** and nothing else. The placeholder never says what kind of thing was removed, because even "health" tells a reader something. - **The original stays where it is.** Your agent never edits, moves or deletes the raw transcript; it writes a new document. - **When unsure, it asks you,** without quoting the passage in its question. For a meeting that's sensitive as a whole, it asks whether to share it at all. ```text Make a sanitized copy of the transcript of [the meeting] for the [name of your shared Space] Space. Keep everything word for word with speakers and times, except replace health, family, personal money, private career, staff or HR matters, credentials and unrelated personal chat with [private discussion redacted]. Don't say what kind of thing you removed. Don't change the original. If you're not sure about a passage, ask me without quoting it. Use the template at https://avalandra.com/help/template-files/cleaned-transcript. Name it with the meeting's short name, the word "transcript" and the meeting date, and nothing longer. ``` Template: [Cleaned transcript](https://avalandra.com/help/templates/cleaned-transcript) Next, have your agent do the follow-up. See [Follow up after a meeting](https://avalandra.com/help/follow-up-after-a-meeting). ### Sanitize every new transcript automatically New transcripts keep arriving, so you can have your agent clean each one as it comes in. Either way, it skips transcripts that already have a cleaned copy, never touches the original, and holds back any transcript it's unsure about until you answer, so nothing private reaches the shared Space by accident. - **On a schedule.** Set a scheduled task in your agent's own tool, such as Claude Code or Codex, to check for new transcripts every hour or so. - **As they arrive.** While your agent is running, it can watch your private Space and clean each new transcript within moments of it landing. ```text From now on, whenever a new meeting transcript arrives in the [name of your private Space] Space, make a sanitized copy in the [name of your shared Space] Space, following https://avalandra.com/help/bring-in-your-meetings.md#sanitize-your-transcripts. Use the template at https://avalandra.com/help/template-files/cleaned-transcript, named with the meeting's short name, the word "transcript" and the meeting date. Check [every hour during working hours, or continuously while you're running]. Skip any transcript that already has a cleaned copy. If you're unsure about a passage, don't publish that transcript; ask me in [where you want questions, such as a Slack direct message] without quoting it, and finish it once I answer. Tell me each time you publish one. ``` ### Using another meeting tool If your meetings happen somewhere else, such as Google Meet, Microsoft Teams or a phone call with notes, your agent can bring them in for you. Most meeting tools let you download a transcript or notes. Give that file to your agent, or tell it where your tool saves them if your agent can reach that place. It adds the full transcript to your private Space with a note of where it came from, then makes a sanitized copy for the shared Space, as above. ```text I use [your meeting tool]. Here's the transcript of [the meeting] from [the date]: [attach it, paste it, or say where it's saved]. Add the full transcript to the [name of your private Space] Space as a document titled with the date, the meeting name and the tool, with a line saying where it came from. Keep the speakers' names and times as they are. Then sanitize it into the [name of your shared Space] Space, following https://avalandra.com/help/bring-in-your-meetings.md#sanitize-your-transcripts. ``` To make it a habit, ask your agent to do this after each meeting, or schedule it in your agent's own tool if it can reach the folder where your transcripts are saved. ### Ask about a meeting ```text What did we decide about [the topic] on [the meeting, such as Tuesday's call with a client]? Tell me who said it, and link me to the transcript. ``` ### If something's off - **Nothing has arrived.** Check that the connection is on and look at its last import. The meeting tool may not have finished processing the meeting yet. - **Only notes, no transcript.** Some meetings only have notes or a summary. Your agent works from whatever came in, and should tell you which it had. - **A meeting arrived twice, or changed.** The meeting tool can update a meeting after it's sent. The connection updates the same document in Avalandra rather than adding a new one. ### Questions ### Which meeting tools bring meetings in automatically? Granola and Zoom. For any other tool, your agent can bring meetings in for you. See [Using another meeting tool](#using-another-meeting-tool). ### Why did a meeting arrive with notes but no transcript? Some meetings only have notes or a summary in the meeting tool, and that's what comes in. Your agent works from whatever arrived and should tell you which it had. ### Can my agent tidy up the folder where meetings arrive? Leave that folder as it is while the connection is on. The connection owns the folder and the meetings in it, and keeps them up to date. Ask your agent to put summaries and notes in a different folder. ### For agents ``` ava connector providers --json ava connector create --file - --idempotency-key KEY ava connector check ID ava connector list ava connector runs ID ``` - `ava connector providers --json` shows what each provider brings in. Granola offers `notes` and `transcripts`. Zoom offers `meeting_transcript`, `recording_transcript` and `summary`. - `create` takes JSON with `display_name`, `scope`, `provider`, `destination`, `interval_minutes`, `sources`, `share_private`, `notifications` and `credentials`. Send secrets only on stdin (`--file -`) or from a file, never as flags and never in a document. Reads never return secrets. - `revise`, `enable`, `pause`, `remove` and `credentials replace` need `--expected-revision`. - The connection owns the documents it imports. Read them; never edit, move, rename or delete them. A document can be revised by the connection, so detect changes by its revision id. - Never copy ids from an example or another organization. Read them from the person's own instance. - To sanitize automatically, either run on a schedule in your host and use `ava changes list --space PRIVATE_SPACE_ID` to find transcripts added since your last run, or stay running with `ava changes watch --space PRIVATE_SPACE_ID` and act on each new transcript as it lands. Key each cleaned copy to its source (such as `granola:` or `ava:`), check for that key before writing, and never publish a transcript with an unanswered question. - When sanitizing, write a new document and never edit the source. Replace every excluded passage with exactly `[private discussion redacted]`, and never name the excluded category anywhere, including in questions, comments, logs or the document title. Ask about uncertain passages without quoting them. Before writing, check the destination's readers can all see what you're putting there. - When you bring in a meeting from another tool, first check the destination folder for a document with the same date and meeting name. Create it with `ava document create` in the private Space the person named, titled like `YYYY-MM-DD - Meeting name (Tool)`, with a first line saying where the transcript came from and who provided it. Keep the transcript as given; put your summary in a separate document. ## Follow up after a meeting > Have your agent write the follow-up from the meeting's sanitized transcript, with a summary for the team, a note for attendees or a client, and the decisions recorded. Source: https://avalandra.com/help/follow-up-after-a-meeting · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/follow-up-after-a-meeting.md and do it with me. Ask me before you decide anything the guide leaves to me, and tell me what you did when you're done. ``` Once a meeting is in Avalandra, your agent can do the follow-up for you. It writes a summary for the team, a note for attendees or a client, a catch-up for anyone who missed it, and records what was decided. It works from the meeting's sanitized transcript, so private discussion stays out. ### Before you start - The meeting is in Avalandra and has a sanitized transcript in the shared Space. See [Bring in your meetings](https://avalandra.com/help/bring-in-your-meetings) and [Sanitize your transcripts](https://avalandra.com/help/bring-in-your-meetings#sanitize-your-transcripts). - You know who each follow-up is for. See [Decide who it's for](#decide-who-it-s-for) below. ### Summarise it for the team A summary gives people who don't need the whole transcript the decisions, actions with owners, quotes that matter and open questions. ```text Summarise [the meeting] into the [name of your shared Space] Space from its sanitized transcript: decisions, actions with owners, quotes that matter and open questions. Use the template at https://avalandra.com/help/template-files/meeting-summary. Name it with the meeting's short name, the word "summary" and the meeting date, and nothing longer. ``` Template: [Meeting summary](https://avalandra.com/help/templates/meeting-summary) ### Send a follow-up to attendees or a client A short note in the reader's terms, covering what you agreed, what happens next, and what you need from them. ```text Write a follow-up for [the person or group, such as the client] after [the meeting], from its sanitized transcript: what we agreed, what happens next with owners and dates, and anything we need from them. Leave out internal discussion. Use the template at https://avalandra.com/help/template-files/client-update. Name it "Update for" followed by their name and today's date, and nothing longer. ``` Template: [Update for a client](https://avalandra.com/help/templates/client-update) If your agent can draft email or post to Slack in its own tool, ask it to send the follow-up there as well, and to show you before it sends. ### Catch up someone who missed it ```text [Teammate's name] missed [the meeting]. From its sanitized transcript and anything else in the [name of your Space] Space since, write them a short catch-up: what we decided, why, and what's waiting on them. Link each point to where it came from. Use the template at https://avalandra.com/help/template-files/catch-up. Name it "Catch-up" followed by today's date, and nothing longer. ``` ### Record what was decided Turn each decision from the meeting into a decision record, so it can be found and explained later. ```text List the decisions made in [the meeting]. For each one I confirm, record it in the [name of your Space] Space with its reason and a link to the moment in the transcript, using the template at https://avalandra.com/help/template-files/decision-record. Name it "Decision" followed by a topic of two or three words, and nothing longer. ``` If a decision replaces an earlier one, see [Update a decision when it changes](https://avalandra.com/help/update-a-decision-when-it-changes). ### Decide who it's for Before your agent writes anything others will read, settle these with it: - **Readers.** Who will see it? Put it in a Space where those people, and only those, are members. - **What stays out.** The sanitized transcript has already lost private discussion. Say if anything else should stay out of this follow-up, such as pricing or internal debate. - **What goes in.** Decisions, next steps, risks, quotes, names, links to the source. Choose the ones you want. - **Links to the source.** Only link to a transcript if every reader can open it. Otherwise the follow-up says where it came from without linking. - **Questions.** When your agent isn't sure, it should ask you one clear question and wait, rather than guess. ### Make it a habit Schedule the follow-up in your agent's own tool, such as a scheduled task in Claude Code or Codex, so it happens after every meeting. Write the request once, with your decisions in it: ```text Every weekday at 5pm, find today's meetings in the [name of your Space] Space that have a sanitized transcript and no summary yet. Summarise each one using the template at https://avalandra.com/help/template-files/meeting-summary, in the [folder name] folder, named with the meeting's short name, the word "summary" and the meeting date. Leave out pricing and anyone's personal details. If you're unsure about something, ask me in [the place you want questions] and skip that meeting until I answer. ``` Each meeting gets one summary, and a second run updates it instead of adding another. On a day with no new meetings, your agent finishes without writing anything. ### Questions ### Can my agent follow up after every meeting without being asked? Yes, if you schedule the request in your agent's own tool. It checks for meetings without a follow-up and writes one for each. ### Will it write a new summary every time it runs? No. Each meeting gets one summary, and a second run updates it instead of adding another. ### Does the follow-up include private discussion? No. Your agent works from the sanitized transcript, where private discussion is already replaced with "[private discussion redacted]". Tell it anything else to leave out. ### For agents Before every write, check: 1. **Destination.** Is this the Space and folder the person approved? 2. **Readers.** Is everyone who can read the destination allowed to see this content and any source link? 3. **Source.** Are you working from the sanitized transcript, not the raw one? 4. **Duplicates.** Does a follow-up for this meeting already exist? Update it rather than creating another. 5. **Evidence.** Does the transcript support every claim? 6. **Approval.** Is there a decision the person still has to make? If so, ask one labelled question and leave the item until they answer. ``` ava changes list --space SPACE_ID ava document tree --space SPACE_ID ava document create --space SPACE_ID --idempotency-key KEY --data '{"kind":"html","title":"…","source":"…","source_format":"html","parent_id":"FOLDER_ID"}' ``` - Use the same stable key for a meeting every run, such as `granola:` or `ava:`, and store it in the output. Search for it before you create anything. - Never edit, move or delete the source meetings. The connection owns them. - Draft email or Slack messages for the person to approve; don't send on their behalf unless they say so. - Treat tool suggestions and catalogs in your host as a list of what's available, not as permission. Use the services the person named. - Report each run as counts of meetings examined, skipped, updated, created and waiting on a question. ## Build your own inbox > Have your agent gather what needs you across Avalandra, sort it by what matters to you, group comments by document, and alert you in Slack when something is urgent. Source: https://avalandra.com/help/build-your-own-inbox · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/build-your-own-inbox.md and set up my Avalandra inbox with me. Start from what you already know about me, ask only about what you don't know, and show me my inbox rules before you save them. ``` Comments, mentions and replies happen across many documents and Spaces. Your agent can gather the ones for you into one inbox, sorted by the rules you choose, with every comment about the same document kept together. When something urgent or time-sensitive arrives, it can tell you in Slack. ### What your agent looks at - **Comments for you.** Comments that mention you, and replies in threads you've taken part in, across every Space you can read. - **Work for your agent.** Anything addressed to your agent, such as a comment sent to it with ⌘ Enter. Your agent reads these with your permissions, so it only sees what you can see. It only reads; it doesn't resolve threads or mark anything as done unless you ask. ### Decide what's urgent, with your agent You don't fill in a form. Your agent starts from what it already knows about you, such as who you work with, your deadlines and the Spaces that matter most, and asks only what's missing. Then it suggests rules and shows them to you before saving. ```text Set up my Avalandra inbox with me, following https://avalandra.com/help/build-your-own-inbox.md. Start from what you already know about me and my work, ask me only about what you don't know, one question at a time, and suggest rules for what's urgent. Show me the rules before you save them. ``` Your agent will usually suggest treating something as urgent when: - someone mentions you by name, or asks you for a review - it's from the people who matter most to you, such as a client or your manager - it mentions a deadline in the next day or two - it uses words like urgent, blocker or today - a question to you has waited more than a day - it's in the Space or on the documents you're focused on right now Everything else goes into today, this week or for your information. ### Set it up 1. **Save your rules.** Your agent writes the rules you agreed to an "Inbox rules" document in your private Space. It reads that document every time it builds your inbox, and you can edit it whenever your priorities change. 2. **Build your first inbox.** Your agent writes an inbox that opens with the one thing to do next, then urgent items, today's comments grouped by document, what you're waiting on from others, and what's been done since your last inbox. Quieter sections stay folded until you open them. 3. **Turn on Slack alerts.** For anything urgent or time-sensitive, your agent sends you a short Slack message with a link. 4. **Keep it current.** Schedule the inbox and the urgent check in your agent's own tool. 5. **Tune it.** After a few days, tell your agent what it put in the wrong place. It updates your rules. Your rules use the [Inbox rules](https://avalandra.com/help/templates/inbox#inbox-rules) layout, part of the Inbox workflow. ```text Using my rules in the "Inbox rules" document in my private Space, build my inbox: everything that needs me across Avalandra, sorted into urgent, today, this week and for your information, with comments on the same document grouped together. Use the template at https://avalandra.com/help/template-files/inbox. Name it "Inbox" followed by today's date, and nothing longer. Put it in my private Space. ``` Template: [Inbox](https://avalandra.com/help/templates/inbox) ### Get alerts in Slack Your agent sends the alert through its own Slack connection. If it doesn't have one yet, connect Slack in your agent's tool first, such as the Slack connector in Claude. ```text Whenever something in Avalandra matches the urgent rules in my "Inbox rules" document, send me a Slack message in [a direct message to me, or the channel to use]. Keep it short: the document, who it's from, why it's urgent, and a link. Don't paste long comment text into Slack. Don't alert me about the same thing twice, and hold anything that isn't urgent between [quiet hours, such as 7pm and 8am]. ``` ### Keep it current Schedule both in your agent's own tool, such as a scheduled task in Claude Code or Codex: ```text Every weekday at [the time you start work], rebuild my Avalandra inbox from my "Inbox rules", as a new inbox document for the day. Every [how often, such as 15 minutes] during [your working hours], check for anything new that matches my urgent rules and alert me in Slack. If nothing is new, do nothing. ``` ### Questions ### Who sends the Slack message? Your agent does, through its own Slack connection. Connect Slack in your agent's tool, then ask it to send alerts as above. ### Can I change what counts as urgent later? Yes. Edit your "Inbox rules" document, or tell your agent what to change and it updates the document. The next inbox uses the new rules. ### Will my agent resolve threads or mark things as read? No. It only reads to build your inbox. Ask it directly if you want it to reply to or resolve something. ### Does it see comments in Spaces I'm not in? No. It works with your permissions, so it only sees Spaces you're a member of. ### For agents - **Find what's for the person.** Get their actor id from `ava comment mention-suggestions --space SPACE_ID` (match their name). In each Space from `ava space list --json`, use `ava changes list --space SPACE_ID` to find documents with recent comment activity, then read threads with `ava document comment list DOC_ID` (HTML) or `ava document text-comment list DOC_ID` (text). Keep threads where a mention targets the person's actor id, or the person wrote an earlier message. - **Find what's for you.** `ava agent check-in --json` lists items addressed to you, the agent. Include them, labelled as work for the agent. - **What else goes in.** "Up next" is the single most important item, with a suggested next step and its deadline. "Waiting on others" is threads where the person asked something and nobody has answered. "Done since your last inbox" is threads resolved or answered since the previous inbox; read the previous inbox document to know what's new, and mark new items "New". Next steps are suggestions; never act on them. - **Sort.** Read the person's "Inbox rules" document first. Put each thread in one tier, either urgent, today, this week or for your information. Group threads under their document, newest first, and give each a short reason ("mentions you", "review requested", "due Thursday"). - **Read only.** Never `ack`, `resolve` or reply while building the inbox. - **Alerts.** Post to Slack only for items newly urgent since the last alert, and remember which you've alerted on. Keep messages to the document title, who, why it's urgent and its `web_url`. Respect quiet hours. - **Rules.** Start from what you already know about the person (memory, instruction files, earlier sessions, who they work with and their deadlines), ask only for what's missing, then write the agreed rules to "Inbox rules" in their private Space with the inbox-rules template, show it to them before saving, and update it when they correct a placement. ## Keep your context golden > Review a Space regularly with your agent so it holds what your team trusts, not everything it ever wrote down. Your agent suggests, you decide. Source: https://avalandra.com/help/keep-your-context-golden · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/keep-your-context-golden.md and do it with me. Ask me before you decide anything the guide leaves to me, and tell me what you did when you're done. ``` Golden context is the part of your record your team trusts, such as decisions that still hold, documents that are current, and the parts of meetings that matter. It doesn't stay that way on its own. Drafts pile up, decisions change, and old copies sit beside new ones. When that happens, an agent can answer from the wrong copy. A short, regular review keeps it trustworthy. Your agent does the looking and suggests what to change. You and your teammates decide. ### What to look for - **Replaced versions.** An old draft next to the one you actually use. - **Copies.** The same content in two places, drifting apart. - **Decisions that changed.** An old decision still reads as if it holds. - **Contradictions.** Two documents that say different things about the same topic. - **Raw material in a shared Space.** Full transcripts or notes where a summary would do. - **Documents nobody has touched in a long time.** Not always wrong, but worth a look. ### Review a Space Ask your agent to review a Space and write up what it suggests, without changing anything yet. A week, the end of a project phase, or before someone new joins are good times. ```text Review the [name of your Space] Space and suggest how to prune it. Look for replaced drafts, copies of the same thing, decisions that have changed, documents that contradict each other, and anything that looks out of date. For each one, say what you'd do (keep it, promote it, archive it, or mark it as replaced) and why. Don't change anything yet. Write it up using the template at https://avalandra.com/help/template-files/pruning-review. Name it "Pruning review" followed by today's date, and nothing longer. ``` Template: [Pruning review](https://avalandra.com/help/templates/pruning-review) ### Then act on it Go through the review with your teammates, decide each item, and tell your agent what to do. The next three guides cover each kind of change: - [Mark the version you trust](https://avalandra.com/help/mark-the-version-you-trust) - [Archive what's been replaced](https://avalandra.com/help/archive-what-has-been-replaced) - [Update a decision when it changes](https://avalandra.com/help/update-a-decision-when-it-changes) ```text Go ahead with the items I've agreed in the pruning review for [name of your Space]: [list the items, or say "everything marked agreed"]. Leave the rest as they are, and tell me what you changed. ``` ### Questions ### How often should we prune? Often enough that nobody has to wonder which copy is current. For an active project that's often weekly. For slower work, at the end of each phase. ### Will my agent delete things on its own? Ask it to suggest first and wait for your decision, as the prompts here do. Prefer archiving over removing, so nothing is lost if you change your mind. ### What if we prune something we still need? Archived documents are still there in their archive folder, and an archived Space can be restored. Before a big edit, your agent can save a named snapshot of a document, and the document can be restored to it later. ### For agents ``` ava document tree --space SPACE_ID ava document status DOC_ID --space SPACE_ID --json ava changes list --space SPACE_ID ``` - `ava document status` returns `standing` (imported, authored, canonical or superseded), `last_edited_at`, `last_reviewed_at` and `canonical_revision_id`. Use them as evidence in the review, and say which you relied on. - Suggest, then wait. Don't move, remove, promote or edit anything until the person agrees to that item. - Prefer archiving to removing. Only run `ava document remove` for documents the person named and explicitly agreed to remove. - After a review, a document the person checked can be marked reviewed with `ava document metadata` (`reviewed` in the JSON body). - If a prompt still contains [square brackets], ask the person what goes there. Never invent a Space, document or folder. ## What you can do with your agent > Everyday things to ask your agent once it's connected, from catching up and recording decisions to writing documents and answering comments. Source: https://avalandra.com/help/what-your-agent-can-do · Commands checked 2026-11-10 ### Have your agent do it A person can paste this to their agent: ```text Read https://avalandra.com/help/what-your-agent-can-do.md. Then look at what's in my Spaces and suggest three of these I'd get the most from this week. ``` Once your agent is connected, you work with Avalandra by asking it in your own words. It reads your team's record, answers with where each answer came from, and writes back what's worth keeping. Everything it writes shows its name, such as "[Your name]'s agent (Claude)". Each request below is something your agent can do today. Copy one, fill in the parts in [square brackets] with your own Space, documents and people, and paste it to your agent. If you leave a bracket in, your agent should ask you what goes there. Claude connected through the connector can do the catch-up, decision, meeting, document and comment requests. Organising Spaces and exporting a Space need an agent connected with the install command, such as Claude Code. When a request produces a document, the prompt points your agent at one of our templates, so what it writes is easy to read in light and dark and on a phone. Each template below is filled with sample content to show the layout; your agent fills it with yours. You can [see every template](#templates) at the end of this guide. ### Catch up Find out what happened while you were busy or away. ```text I've been away since [date]. Read the [name of your Space] Space and write me a catch-up: what we decided, why, and what's waiting on me. Link each point to where it came from. Use the template at https://avalandra.com/help/template-files/catch-up. Name it "Catch-up" followed by today's date, and nothing longer. ``` Template: [Catch-up](https://avalandra.com/help/templates/catch-up) ```text What's waiting on me in Avalandra? List the comments that mention me and the replies to my comments, newest first. Use the template at https://avalandra.com/help/template-files/waiting-on-you. Name it "Waiting on you" followed by today's date, and nothing longer. ``` Template: [Waiting on you](https://avalandra.com/help/templates/waiting-on-you) ### Ask about a decision Get the answer from the record, with who decided and the source. ```text Why did we decide [the decision you're asking about]? Look in the [name of your Space] Space, tell me who decided and when, and link me to where it was decided. If I ask you to save the answer, use the template at https://avalandra.com/help/template-files/decision-explainer. Name it "Why" followed by a topic of a few words, and nothing longer. ``` Template: [Why it was decided](https://avalandra.com/help/templates/decision-explainer) ### Record a decision Keep the reason with the decision, so nobody has to remember it. ```text Record a decision in the [name of your Space] Space: [what was decided], because [the reason]. Link it to [where it was decided, such as today's meeting]. Use the template at https://avalandra.com/help/template-files/decision-record. Name it "Decision" followed by a topic of two or three words, and nothing longer. ``` Template: [Decision record](https://avalandra.com/help/templates/decision-record) ### Ask about a meeting Meetings come in from Granola and Zoom once they're connected. See [Bring in your meetings](https://avalandra.com/help/bring-in-your-meetings). ```text What did [person] say about [topic] on [the meeting]? Quote their words, say who else spoke, and link me to the transcript. ``` ```text Summarise [the meeting] into the [name of your Space] Space from its sanitized transcript: decisions, actions with owners, and open questions. Use the template at https://avalandra.com/help/template-files/meeting-summary. Name it with the meeting's short name, the word "summary" and the meeting date, and nothing longer. ``` Template: [Meeting summary](https://avalandra.com/help/templates/meeting-summary) ### Write and update documents Your agent can write a document into a Space or update one that's already there. ```text Write a one-page brief for [what it's about] from what's in the [name of your Space] Space, and put it in the [folder name] folder. Use the template at https://avalandra.com/help/template-files/brief. Name it with a topic of two or three words followed by "brief", and nothing longer. Tell me the link when it's done. ``` Template: [Brief](https://avalandra.com/help/templates/brief) ```text Update [the document] in the [name of your Space] Space: [the changes you want]. Keep the rest as it is. ``` ### Comment and reply Your agent can comment on a document, reply in a thread and mention a teammate. Comments are plain text, so there's no template. Ask for a short reply that answers first and links its source. ```text Read [teammate]'s comment on [the document] and reply with what we decided about [the topic]. Answer in one or two sentences, link the decision, and mention [anyone who should see it]. ``` ### Follow up after a meeting Turn a meeting into an update for people who weren't there. See [Follow up after a meeting](https://avalandra.com/help/follow-up-after-a-meeting). ```text Write a short update for [the person, such as a client] from this week's work in the [name of your Space] Space: what's done, what's next and anything we need from them. Leave out internal discussion. Use the template at https://avalandra.com/help/template-files/client-update. Name it "Update for" followed by their name and today's date, and nothing longer. ``` Template: [Update for a client](https://avalandra.com/help/templates/client-update) ### Organise your Spaces Move documents to where they belong, and share a Space with a teammate who already has an account. ```text Move [the documents] in the [name of your Space] Space into the [folder name] folder, and add [teammate] so they can edit the Space. Then write down what changed, using the template at https://avalandra.com/help/template-files/space-changes. Name it "Space changes" followed by today's date, and nothing longer. ``` Template: [Space changes](https://avalandra.com/help/templates/space-changes) To bring someone new into your team's Avalandra, an admin invites them by email and they accept in their browser. Your agent can't accept an invite for anyone. ### Templates Each template is a ready-made Avalandra document filled with sample content. Your agent swaps in your content and keeps the layout. - [Catch-up](https://avalandra.com/help/templates/catch-up): what changed while someone was away - [Waiting on you](https://avalandra.com/help/templates/waiting-on-you): mentions, replies and review requests - [Why it was decided](https://avalandra.com/help/templates/decision-explainer): the answer, who decided, the reasons and sources - [Decision record](https://avalandra.com/help/templates/decision-record): one decision with its reason and source - [Cleaned transcript](https://avalandra.com/help/templates/cleaned-transcript): a full transcript with private discussion taken out - [Meeting summary](https://avalandra.com/help/templates/meeting-summary): decisions, actions, quotes and open questions - [Brief](https://avalandra.com/help/templates/brief): a new working document, such as a brief or spec - [Update for a client](https://avalandra.com/help/templates/client-update): what's done, what's next, what you need - [Space changes](https://avalandra.com/help/templates/space-changes): what moved and who was added - [Export record](https://avalandra.com/help/templates/export-record): when a Space was exported and where the file is ### Questions ### Who can see what my agent writes? Everyone who's a member of the Space it writes in. Each change and comment shows your agent's name. Anything in your private Space is visible only to you. ### Does Avalandra remember what I said to my agent? Only what gets written into a Space. Your conversation stays in your agent's own tool. When something's worth keeping, such as a decision and its reason, ask your agent to record it in the Space. ### Can my agent accept an invite for me? No. Invites are for people. You open the link in the email and sign in in your browser. Agents connect with their own one-time code instead. ### Can I move a document to another Space? Yes. Ask your agent to move it. Its link keeps working after the move. ### Can I take my data out? Yes. Ask your agent to export the Space, with its history, to a ZIP file on your computer, and it does the whole thing for you. A large Space can take a few minutes. The file isn't encrypted, so keep it somewhere only the right people can open. This needs an agent connected with the install command, such as Claude Code. Your agent can also note the export in the Space using the [Export record](https://avalandra.com/help/templates/export-record) template. ### Can I bring an export back in? Yes, into a new, empty Space. An export doesn't merge into a Space that already has things in it. ### For agents What each request above uses: - **Read and answer:** `ava document tree --space SPACE_ID`, `ava document get DOC --space SPACE_ID --json`, `ava document render DOC`, and `ava changes list --space SPACE_ID` for what changed. Cite records with the `ref` the CLI returns, and give people `web_url`. - **What's waiting:** `ava agent check-in --json`. It is read-only and lists mentions, replies and review requests. - **Write a document:** `ava document create` with `"kind":"html"` and a `parent_id` for the folder. Replace an HTML document's source with `ava document edit DOC --expected-revision REV`. - **Comment:** `ava document comment create|reply` on HTML documents and `ava document text-comment create|reply` on text documents. A mention needs `@Name` in the body and an entry in `mentions` from `ava comment mention-suggestions`. - **Organise:** `ava document move` within a Space, `ava document transfer` between Spaces. Re-read `ava document tree` afterwards. - **Share a Space:** `ava space members grant --space SPACE_ID` with the person's `principal_id` from `ava user list`. Only an organization admin can create invites, and only a person can accept one. - **Export:** `ava space export SPACE_ID --output PATH --idempotency-key KEY`. It waits for the archive, checks it and writes it with owner-only permissions; a large Space takes minutes (a 126 MB Space took about two). The response's `verified` should be `true`, and `security` reads `unencrypted`, so tell the person. `ava space exports SPACE_ID` lists past exports. - **Templates:** fetch the template the person's prompt names, and title the document with the template's short name pattern (on its page and in the file's top comment), such as "Catch-up 10 Nov", never a long descriptive title, and don't repeat the title inside the document. Keep its `