SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI developer platformsbeever-atlas

beever-atlas

Team knowledge base turning Slack, Discord, Teams and Mattermost chats into a knowledge graph and wiki with cited answers, semantic search and expert finding.

Open website

About this tool

Beever Atlas is an open-source conversational knowledge base for teams, with the goal of automatically turning a team's existing chat content in Slack, Discord, Microsoft Teams, and Mattermost into a self-maintaining wiki. It extracts atomic facts from conversations, deduplicates them, and clusters them into topic pages with citations; at the same time, it uses graph storage to link people, decisions, and projects mentioned across channels. Users can ask questions in natural language, and answers cite back to the source messages, usable both through the dashboard and through MCP integration with Claude Code and Cursor.

Main capabilities

  • Multi-platform integration: supports conversation retrieval and synchronization for Slack, Discord, Microsoft Teams, and Mattermost.
  • Automatic wiki generation: each synchronized channel generates a continuously updated wiki, including sections such as topics, entities, decisions, open questions, and timelines, and it is incrementally rebuilt as new messages arrive.
  • Dual memory system: a three-layer semantic store (channel/topic/atomic fact) for fast hybrid search, and a graph store for extracting entities and their relationships.
  • Q&A with citations: the Q&A agent first retrieves from the wiki, and falls back to raw messages for facts that have not yet been distilled; answers can be traced back to the source messages.
  • MCP server: exposes 28 tools at /mcp for external AI agents such as Claude Code and Cursor to query the team knowledge base, and also provides a standalone stdio mode.
  • Media persistence: images, PDFs, and videos in synchronized channels are persistently stored, by default in the database, and MinIO/S3 can also be used.

How to use

The project is released as a Docker Compose stack, including three services: backend, bot, and frontend, as well as four data stores: Weaviate, Neo4j, MongoDB, and Redis.

  1. Get the code.
  2. Optional: run make demo to start the full stack with preloaded public Wikipedia corpus (Ada Lovelace and Python history); the seed data uses precomputed fixtures and requires no API key; asking questions via /api/ask requires a free GOOGLE_API_KEY because the Q&A agent calls Gemini.
  3. Before installation, two free keys are required, both offering free quotas.
  4. Choose a deployment method:
    • One-line install (recommended):./atlas guides you through a five-step checklist, including embedding model provider (Jina / OpenAI / Cohere / Voyage / Gemini / Mistral / Ollama), agent LLM provider (Google Gemini / OpenAI / Anthropic / Mistral / DeepSeek / Groq / MiniMax / Ollama / custom), graph backend (Neo4j or skip), optional integrations (Tavily web search, MCP server), and authentication token. The installer checks docker and docker compose, copies .env.example to .env, automatically generates CREDENTIAL_MASTER_KEY and WEAVIATE_API_KEY, performs a port conflict precheck, starts the stack, and polls /api/health.
    • Manual Docker: step-by-step control, fill in the two required keys in .env and generate the two required keys.
    • Local development: databases run in Docker, application services run natively to support hot reload, requiring Python 3.12+ and uv, Node.js 20+, access http://localhost:5173。
  5. Open the dashboard: options 1 and 2 access http://localhost:3000，选项 3 accesses http://localhost:5173。
  6. Sync channels: In the dashboard, go to Connections → Add Workspace → select a channel → Sync, or do it via the API.

In real mode (default, ADAPTER_MOCK=false), connect workspaces under Settings → Connections; Slack / Discord / Teams tokens are entered through the UI rather than .env. Mock mode (ADAPTER_MOCK=true) uses fixture data and is suitable for local UI iteration without platform credentials.

When to use it

  • Teams want the knowledge base to grow automatically from existing chats rather than being maintained manually.
  • Q&A with traceable citations is needed, where answers can link back to the original Slack/Discord/Teams threads.
  • When new members join a channel, they can read a distilled wiki instead of scrolling through months of history.
  • You want external AI agents such as Claude Code and Cursor to query the team knowledge base via MCP.
  • Graph-aware reasoning is needed to answer relational questions like "who worked with whom on X."

Limitations and notes

  • All /api/* endpoints are marked unstable in 0.1.0; v0.2.0 will introduce the /api/v1/* prefix, and clients pinned to the current paths will break.
  • Telemetry is not collected by default; all LLM calls use API keys configured by the user, and data stays in a database controlled by the user.
  • Before production deployment, rotate the placeholder keys in .env.example and switch the environment flags.
  • The project uses the Apache License 2.0; see NOTICE for third-party attributions.
  • Community channels include Discord, X/Twitter, the official website, and GitHub Discussions; for commercial support, partnerships, or media inquiries, contact tech@beever.ai.

Related tools

Browse category