How aweb.ai runs its company with AI agents.
This public repo holds the product docs, agent instructions, publishing pipeline, and the homes for the permanent agent team. Sensitive data (outreach contacts, identity keys) lives in a separate private repo (co.aweb).
Start here:
| Document | What it covers |
|---|---|
| docs/team.md | Permanent responsibility areas and task-scoped work pairs |
| docs/invariants.md | Guiding principles for every design decision |
| docs/user-journey.md | What users experience at each stage |
| docs/value-proposition.md | Why aweb exists, who it's for |
| docs/agent-first-company.md | How aweb.ai runs through areas, work pairs, and feedback loops |
| docs/company-dashboard.md | Queryable company dashboard and signal inventory |
| status/ | Current focus and state |
For specific topics:
| Document | What it covers |
|---|---|
| docs/audiences.md | Who uses aweb, what they want, where they are |
| docs/capabilities.md | What aweb provides, mapped to user journey stages |
| docs/decisions.md | Decision log with commit hashes |
| docs/aweb-high-level.md | Identity and protocol architecture |
| docs/strategy.md | Go-to-market playbook |
| docs/atext-spec.md | Product spec for atext.ai (not yet started) |
Permanent agents run from subdirectories under agents/. Each has
an AGENTS.md (instructions) and handoff.md (state for continuity).
The organization is responsibility-area based: each agent owns a work surface and the artifacts that keep it legible.
| Agent | Responsibility area | Directory |
|---|---|---|
| Randy | Engineering | agents/engineering |
| Avi | Direction | agents/direction |
| Charlene | Outreach | agents/outreach |
| Amy | Support | agents/support |
| Enoch | Operations | agents/operations |
| TBD | Analytics | agents/analytics |
Significant repo work uses task-scoped builder/reviewer agents created
with aw workspace add-worktree, not permanent repo-manager agents.
See docs/team.md for how they work together.
Outreach owns the publishing pipeline:
| File | What |
|---|---|
publishing/plan.md |
Content calendar — what to write, where, when |
publishing/voice.md |
How we talk |
publishing/landscape.md |
Agent-to-agent ecosystem map |
publishing/history.md |
What we published and where |
publishing/drafts/ |
Blog posts, video scripts |
| Directory | What it holds |
|---|---|
status/ |
Status files maintained by agents (engineering, product, outreach, weekly) |
Sensitive operational data lives in a separate private repo:
- Outreach contacts, engagement history, monitoring targets
- Competitive positioning
- Identity key backups
Agent instructions live in AGENTS.md files (root and per-agent).
CLAUDE.md is a symlink to AGENTS.md so Claude Code picks it up
automatically. Non-Claude agents read AGENTS.md directly.