| name | agentdb233 |
| description | Use when Codex should share project knowledge, norms, code context, Git refs, skills, MCP endpoints, or agent memory through a single agentdb233 external brain instead of reading and maintaining separate per-agent documentation. Trigger for agentdb233, shared AI brain, context pack, build/search project index, import AGENTS.md/CLAUDE.md/.clinerules norms, or connect Codex/Claude/Cline/OpenCode to one common knowledge base. |
agentdb233
Use one shared brain. Do not maintain separate Codex/Claude/Cline/OpenCode docs when agentdb233 is available.
Quick Workflow
- Ensure server exists:
agentdb233-server status
- Build or refresh project index:
agentdb233-server start
curl -X POST http://127.0.0.1:23390/api/index/build \
-H 'content-type: application/json' \
-d '{"project":"PROJECT","repo":"REPO","tracked_only":true,"incremental":true}'
- Import shared norms once:
curl -X POST http://127.0.0.1:23390/api/norms/import \
-H 'content-type: application/json' \
-d '{"project":"PROJECT","repo":"REPO"}'
- Retrieve compact context:
curl 'http://127.0.0.1:23390/api/context/pack?project=PROJECT&q=QUERY&budget=8000'
MCP
Prefer MCP when host supports it:
agentdb233-server mcp
Tools:
build_index: build shared repo index.
search_context: return compact context pack.
git_refs: list branches/tags.
Rules
- Use
tracked_only=true for normal code repos.
- Use
incremental=true after first index.
- Use
ref when context must match a branch/tag/commit without switching worktree.
- Prefer
/api/context/pack over reading whole docs.
- Add durable facts to
/api/knowledge; do not scatter new norms across agent-specific files.
- Keep per-agent files as thin pointers to agentdb233 when possible.