Connect from Claude
Most agent setups treat MCP as the memory: every turn, the agent calls GitHub or Slack tools to reconstruct what happened. That knowledge stays siloed in the session.
Lobu runs a data pipeline instead. Connectors poll and webhooks push into one append-only org log; watchers and chat agents share the same knowledge graph. MCP here is for recall and write — ingestion still flows through connectors and webhooks.
Give Claude durable, structured memory it can search and append to, so the same recall is available across Claude, ChatGPT, and your own agents.
Lobu is open-source but we also have a managed cloud. If you'd like to try, sign in at app.lobu.ai, then point Claude at https://app.lobu.ai/mcp as the MCP endpoint. The same URL works for every workspace. Pick one below to scope it to that example.
Paste this into Claude and it will install Lobu for you.
Connect Claude
- Open Settings → Connectors → Add Custom Connector in Claude Desktop or claude.ai.
- Paste https://app.lobu.ai/mcp as the MCP URL.
- Complete the Lobu sign-in flow.
- Enable the connector and choose the workspace Claude should use.
For Claude Code, run claude mcp add --transport http lobu https://app.lobu.ai/mcp instead and complete the OAuth flow when prompted. If you are self-hosting, replace https://app.lobu.ai/mcp with your custom instance URL (e.g., http://localhost:8787/mcp).
Claude Code and Claude Desktop
Claude Code uses the same MCP endpoint, registered through `claude mcp add`. The OAuth flow is handled in your browser the first time you call a memory tool.
Pair the connector with a project-level instruction file (or a small skill) that tells Claude when to search memory before answering and when to save what it just learned.