Shared ownership
A teammate lives on a shared branch, so its instructions, memory, and history belong to the team, not one person’s setup.
Learn moreBuild on what your team teaches them
Raise AI teammates your whole team can teach
You shouldn’t have to start over every time. Give teammates memory, teach them skills, connect them to your tools, and bring them where your team works. People decide their scope and when they ask for help, and anyone can pick up where a colleague left off. What your most AI-enabled people figure out becomes something the whole team can build on.
Shared ownershipMemory & knowledgeIn your channelsSchedulesSkills & toolsTeach by talkingIdentity & boundaries
A teammate lives on a shared branch, so its instructions, memory, and history belong to the team, not one person’s setup.
Learn moreMemory & knowledge
Teammates keep memory across conversations, and Knowledge gives people and agents one shared place for runbooks and decisions.
Learn moreIn your channels
Mention a teammate in Slack, Discord, GitHub, or Shortcut. It replies where you asked, running as the person who asked.
Learn moreSchedules
Standups, digests, and audits run on a schedule, each leaving a full transcript your team can review.
Learn moreSkills & tools
Package the steps your team repeats as skills, and connect teammates to your tools through MCP.
Learn moreTeach by talking
Onboard a teammate through a guided first conversation, then keep teaching it by correcting it, like a new colleague.
Learn moreIdentity & boundaries
Give each teammate a clear job, a voice, and limits: what it does on its own and when it stops to ask.
Learn moreStop starting over
Pick up where you left off
Good context shouldn’t stay buried in an old conversation. A teammate keeps plain-markdown memory on its own branch: who it is, what it knows about the people it works with, and a daily journal of decisions and learnings. Each new session starts by reading it.
For context the whole team needs, Agor Knowledge gives people and teammates one shared place for runbooks, decisions, and prompts. Memory is only as good as what gets written down, so your team can read it, edit it, and correct it.
- Identity and a daily journalSOUL.md, IDENTITY.md, USER.md, and dated memory files, versioned in git alongside the teammate.
- Shared KnowledgeMarkdown documents in namespaces, with version history, links between pages, and text search. Semantic search is available when an admin turns on embeddings.
Where your team works
Bring it where your team already talks
Not everyone needs to live on the board. Mention a teammate in Slack or Discord, or tag it on a GitHub PR or Shortcut story, and Agor starts a session on the right branch and replies where you asked. Microsoft Teams is coming soon to Agor Cloud.
When a sender is matched to an Agor user, the session runs as that person, with their access and their name on the record. Each channel has its own model, tools, and approval rules.
- Mention to startIn Slack and Discord channels, a teammate only answers when someone mentions it.
- You decide how much ropeKeep a person approving each tool call from the linked Agor session in one channel, and let well-understood work run on its own in another.
Message gateway and its security notes Meet Blake, our deal desk agent
Recurring work
Standups, digests, and audits on a schedule
Some work should happen every morning whether or not anyone remembers to ask. Attach a cron schedule to a teammate’s branch, and each run starts a fresh session with the branch and board context it needs.
A teammate’s heartbeat is a schedule that reads HEARTBEAT.md: check in-progress work, triage what’s new, write the day’s notes to memory. Every run leaves a full transcript your team can review.
- Any cadenceFrom every few minutes to once a day, set with a visual cron builder. Schedules run in UTC.
- No pile-upsRuns are deduplicated, and after downtime only the most recent missed run fires.
- Pick the agentChoose Claude Code, Codex, Gemini, or another supported agent for each schedule.
Teach skills, connect tools
Teach it the way your team works
Package the steps your team repeats as skills: a release check, a review checklist, a weekly report. Skills live with the teammate, so the method is written down instead of living in one person’s head.
Then connect the systems it needs. Skills and MCP servers let a teammate work with tools like Slack, GitHub, Linear, and Datadog, and Agor’s own MCP server lets it create branches, spawn sessions, and organize boards.
- SkillsRepeatable methods for SaaS tools, command-line tools, and APIs, defined alongside the teammate.
- MCP CatalogReviewed remote servers you connect once, with OAuth or a key, then attach to sessions.
- Agor’s own MCPThrough Agor’s built-in MCP server, a teammate works with the same API you do: spawning sessions, moving branches, scheduling runs, all within your permissions.
Teach it by talking
Onboard it like a new teammate
You don’t need a perfect system prompt on day one. A new teammate starts with a guided first conversation that works toward a real outcome for you, while it learns your goals, your context, and how you like to work.
From there, teaching is ordinary conversation. Correct a weak answer, point it at the right doc, or tighten how it reports, and it keeps what it learns in its memory for next time.
- Start from a personaPick a persona from the gallery or start blank, then give your teammate a name and its own board.
- Onboarding with a goalONBOARDING.md steers early sessions toward something useful, not a configuration checklist.
- Keep correctingReview its output, correct it, and fold the fix into its runbook. That loop is how it gets better.
Identity and boundaries
A clear job and clear limits
A useful teammate fits in one sentence: it helps this team handle this kind of work across these systems. Its identity files hold that purpose, along with its voice and the principles it works by. Those files come from agor-teammate, the public framework every Agor teammate is built on, which borrows its core ideas from OpenClaw.
People stay in charge of scope. Decide what it can do read-only, what it drafts for review, and when it should stop and ask. Start with a tight loop, and widen its autonomy as it earns trust.
- Purpose and voiceIDENTITY.md and SOUL.md describe who it is, who it serves, and how it communicates.
- Approval boundariesPermission modes decide whether a person approves each tool call or the teammate works on its own.
- Access that fits the roleScope its tools and Knowledge to the job, not to everything the company has.
Agent modeling 101 The agor-teammate framework Agor and OpenClaw
Go deeper in the docs
Explore more of Agor
Bring your team and agents together
Start on your own with Agor Community Edition and bring colleagues in as you go, or talk to us about rolling Agor out across your team. Agor Cloud is here when you’d rather we run it.