agor-teammate

Framework for AI teammates that live inside Agor.

This repo is a template. Clone it, give it to an AI agent, and you get an AI teammate with a filesystem home base plus a Knowledge Base-first working model — orchestrating work through the Agor multiplayer canvas.

Inspired by OpenClaw, adapted for Agor's branch/session/board primitives and Agor Knowledge Base.

---

How it works

The framework has two layers:

1. Branch/filesystem home base in git — core framework files, identity, operating manual, executable skills, scripts, local data, and code/file workbench.
2. Agor Knowledge Base — long-term, user-visible memory/docs: notes, plans, designs, decisions, drafts, reference material, and shareable artifacts. It is searchable, versioned, permissioned, linkable, and graph-aware.

On each fresh session, a teammate reads its core local files, then searches/reads its assigned Knowledge namespace for current context. The teammate files useful memory back into Knowledge via Agor MCP, especially with the teammate memory append tool.

---

Local files at a glance

File	Purpose
AGENTS.md (= CLAUDE.md)	Always-loaded operating manual; points to Knowledge-first model
KNOWLEDGE.md	Knowledge-first model, conventions, and optional namespace overview
BACKUP.md	Git backup model for teammate home/base files
ONBOARDING.md	Live first-run guide and progress record (deleted after)
BOOT.md	Startup checklist
SOUL.md	Values and communication style
IDENTITY.md	Name, vibe, board config, primary Knowledge namespace
USER.md	Minimal profile of the human
BOARD.md	Agor board zones + workflow
HEARTBEAT.md	Optional periodic tasks
TOOLS.md	Env-specific shortcuts (incl. roster of repos)
skills/	Local onboarding/emergency procedures, including connect-saas and agor-gateway-channels; prefer Knowledge for evolving/shareable skills

---

Onboarding defaults

First run starts with one plain question about the user, identifies where their real work lives, and moves quickly from any useful connection to a result based on live context. The teammate offers to configure Agor while showing the user where the same configuration can be reviewed or changed. ONBOARDING.md keeps a short live progress record across the first interactions. After proving value, the teammate may offer to repeat it on a useful cadence, meet the user in their existing channel, or share it with the people who need it. Identity, sharing preferences, and backup are introduced only when relevant.

Knowledge-first defaults

Use Agor Knowledge for:

• Daily memory and decisions
• Plans, design docs, notes, meeting notes, and research
• Drafts and shareable user-facing docs
• Long-lived reference material and reusable skills/procedures
• Internal graph links between concepts/docs via agor://

Use the local filesystem for:

• Core framework/identity/safety files
• KNOWLEDGE.md high-level model and optional namespace overview
• Repo-native scripts/config/templates, executable skills, data files, and test fixtures
• Temporary materialized Knowledge docs while editing locally

See KNOWLEDGE.md for the detailed decision table and conventions.

---

Branch model

main                       ← this template (framework only)
private-coachbot           ← one running teammate's home/base
private-saul               ← another
kb-first-teammate          ← yours

Each teammate lives on its own branch. Git backs up the teammate home/base files; Agor Knowledge stores long-term memory and docs. See BACKUP.md.

• Running teammates never PR their personal branch anywhere.
• Running teammates never fork the public repo for private state.
• Framework improvements come as PRs against main from contributors/clean branches.

---

Getting started

1. Clone or import this repo into your Agor setup.
2. Create a branch for your teammate: git checkout -b <your-teammate-name>.
3. Assign or identify the teammate's primary Agor Knowledge namespace.
4. Start an Agor session in the branch.
5. The teammate follows ONBOARDING.md across its first interactions — understand a goal, orient the user, connect useful context when approved, and produce a concrete result.
6. After value is established, the teammate suggests a backup setup for the teammate home/base files (private repo, if needed).

---

Resources

• Agor: agor.live
• OpenClaw (inspiration): openclaw.ai
• Connector skill: skills/connect-saas.md
• Gateway channel skill: skills/agor-gateway-channels.md