Open Agent Tag is an agentic engineering assistant for teams that want one Slack-native agent to reason across code, pull requests, product docs, design files, observability data, bug reports, and release workflows.
It keeps the original integration surface intact:
- Slack app and Socket Mode listener
- FastAPI REST API with SSE streaming
- Claude-based tool-calling loop and specialized sub-agents
- GitHub repository, pull request, workflow, review, and worktree tooling
- Confluence, Figma, Lucid Charts, Grafana/Loki, Jam.dev, and Slack context tools
- PostgreSQL-backed conversations, sessions, memories, review metadata, and user preferences
- Optional OpenAI-compatible/NVIDIA NIM proxy support
- Coding workflows through isolated git worktrees
- Release note generation with optional Confluence publishing
- Python 3.11+
- PostgreSQL
- Slack app credentials
- Anthropic API key
- GitHub token or GitHub App credentials
- Optional credentials for Confluence, Figma, Lucid, Grafana, Jam.dev, Databricks/Lakebase, and private package registries
uv sync --all-extras --devOr install in editable mode:
pip install -e .Copy the sample environment file and fill in the credentials you need:
cp .env.example .envEdit config/repos.yaml to describe the repositories Open Agent Tag should index, review, and operate on. Edit config/slack.yaml to describe Slack channels that should guide workspace search.
uv run open-agent-tagRun without Slack by leaving SLACK_BOT_TOKEN and SLACK_APP_TOKEN empty. The REST API still starts and serves /api/v1/chat when database settings are valid.
The chat endpoint streams responses as Server-Sent Events:
curl -N -X POST http://localhost:8000/api/v1/chat \
-H "Content-Type: application/json" \
-H "Authorization: Bearer <user-api-key>" \
-d '{"message": "Where is the login API?", "conversation_id": "optional-id"}'Useful endpoints:
| Method | Endpoint | Description |
|---|---|---|
| POST | /api/v1/chat |
Streaming chat |
| GET | /api/v1/conversations |
List conversations |
| POST | /api/v1/conversations |
Create a conversation |
| GET | /api/v1/conversations/{id} |
Get a conversation |
| DELETE | /api/v1/conversations/{id} |
Delete a conversation |
| GET | /api/v1/conversations/{id}/messages |
List conversation messages |
| GET | /api/v1/users/me |
Current user profile |
| PATCH | /api/v1/users/me |
Update current user profile |
Open Agent Tag exposes tools through a compact MCP server with list_tools, call_tool, process_large_result, ask_user_question, and todo_write.
Tool categories include:
- Code search and file reading
- Pull request inspection and review actions
- Confluence page search and image fetching
- Figma file browsing
- Lucid document search and diagram export
- Grafana/Loki log and dashboard search
- Slack search, thread reading, links, files, and images
- Jam.dev recording, console, network, screenshot, event, and transcript lookup
- Persistent memory search and storage
- Git worktree, commit, pull, merge, status, and PR creation
open-agent-tag/
├── src/open_agent_tag/
│ ├── agents/ # Claude SDK agent definitions, runner, sessions, hooks
│ ├── api/ # FastAPI routes, auth, middleware, schemas, streaming
│ ├── common/ # Settings, context, flags, cost tracking, formatting
│ ├── db/ # SQLAlchemy models, migrations, repositories, engine
│ ├── handlers/ # Slack event, action, delivery, question, stream handlers
│ ├── infra/ # Health, telemetry, tracing, webhook handling
│ ├── proxy/ # OpenAI-compatible proxy and provider adapters
│ ├── services/ # Integrations and domain services
│ ├── tools/ # MCP server, dispatcher, schemas, truncation, user prompts
│ └── main.py # Application bootstrap
├── config/ # Example repo and Slack configuration
├── docker/ # Container helper scripts
├── scripts/ # Maintenance/setup scripts
├── tests/ # Unit, integration, e2e, and eval tests
├── .env.example
├── Dockerfile
├── pyproject.toml
└── README.md
uv run ruff check .
uv run pytest -m "not e2e and not eval"E2E and eval tests call live services. Run them only after supplying the matching credentials.
MIT