Beever Atlas is an open-source Python project that transforms team conversations from Slack, Discord, Microsoft Teams, and Mattermost into a searchable knowledge base. Hosted on GitHub at Beever-AI/beever-atlas with 282 stars, it processes chat history to extract atomic facts, remove duplicates, and organize them into topic pages complete with citations. A graph store connects mentions of people, decisions, and projects across channels. Users query the system in natural language for answers tied back to original messages, either through a dashboard or integration with tools like Claude Code and Cursor via MCP.

The tool addresses the challenge of scattered team knowledge in chat apps. Instead of manual wiki updates, it automates the process: sync messages, distill them via a multi-stage pipeline, and maintain an LLM-powered wiki that evolves with new chats. Built with Google's Agent Development Kit (ADK) and released under the Apache 2.0 license, it suits teams already active in these platforms who want a low-effort way to preserve institutional memory.

Core features

Beever Atlas handles ingestion and querying across multiple steps, as shown in its demo clips:

  • Multi-platform support: Connects to Slack, Discord, Teams, Mattermost workspaces, or import files. A single bot manages all integrations.
  • Message syncing: Pulls channel history on demand or via schedule. The process is resumable and respects platform rate limits.
  • Memory ingestion pipeline: Uses a six-stage ADK workflow to break down messages into facts, entities, and relationships.
  • LLM-generated wiki: Creates per-channel pages covering overviews, topics, people, decisions, and full citations. Pages update automatically as new data arrives.
  • Query agent: Answers natural language questions with streamed responses and source links, accessible via dashboard or external AI tools.

These features run through a web dashboard for browsing and management. The project includes a Discord community for support (discord.gg/VshBCUUX) and links to Beever.ai for related updates.

Getting it running

As a Python project, Beever Atlas requires a standard development setup. Clone the repository from GitHub:

git clone https://github.com/Beever-AI/beever-atlas.git
cd beever-atlas

Detailed installation steps, dependencies, and configuration appear in the official documentation at docs.beever.ai/atlas. Expect to set up a Python environment (version unspecified in the repo overview, but typical for ADK-based tools), install required packages via pip, and configure API keys for chat platforms and any LLMs used in processing.

Once installed, connect your workspaces through the dashboard. Sync starts with selecting channels, then the pipeline ingests data. The web interface (under web/public) provides the entry point, with assets like logos and banners included for theming. For production, deploy behind a reverse proxy, as it's designed for self-hosting. Rate-limit awareness in syncing helps avoid disruptions during initial pulls of large histories.

Who this is for

Teams reliant on chat apps for daily coordination benefit most. Product managers tracking decisions scattered across threads, engineers querying past bug fixes mentioned in Discord, or support teams recapturing client interactions from Teams—all find value in the auto-built wiki. Small to mid-sized groups (under 100 members per workspace) see quicker wins, as sync times scale with message volume.

File imports extend it to non-chat data, like exported logs. If your workflow involves frequent "where did we decide that?" searches, the citation-backed answers reduce time lost to scrolling histories. Remote or async teams, where chats serve as the single source of truth, get a persistent layer without extra documentation overhead.

How it compares

Beever Atlas stands out for its LLM-first approach to wiki building, unlike traditional tools like Notion or Confluence that rely on manual page creation. Open-source alternatives such as Logseq or Obsidian focus on personal notes rather than multi-platform chat ingestion. For chat-specific knowledge extraction, projects like Memex or custom RAG setups with LangChain exist, but they lack the integrated graph store and ADK pipeline here.

It's heavier on LLM dependencies than lightweight scrapers like slack-archive exporters, which dump raw JSON without processing. Commercial options from Beever.ai's ecosystem offer hosted versions, but this GitHub repo provides the self-hosted core. If you need broader search federation (e.g., including email or docs), tools like Haystack or Weaviate might pair better as backends.

Limitations and fit

Solo users or massive enterprises might skip it—individual setups rarely justify chat-to-wiki automation, while large orgs face compliance hurdles with chat exports. The six-stage pipeline demands compute for heavy syncing, so test on modest hardware first.

Check the full README and docs at github.com/Beever-AI/beever-atlas or docs.beever.ai/atlas for latest details.

View on GitHub →Official Website →