Installation¶
System Requirements¶
| Component | Minimum | Notes |
|---|---|---|
| Python | 3.12+ | Required since v0.4.0 |
| RAM | 512 MB | 1 GB+ for large knowledge bases |
| Disk | 200 MB | Base install; data grows with your KB |
| OS | Linux, macOS, Windows (WSL2) | Docker image is linux/amd64 and linux/arm64 |
Models downloaded on first use
Lithos uses a sentence-transformers model (all-MiniLM-L6-v2, ~90 MB) for semantic search and the spaCy en_core_web_sm model for entity extraction. PyPI installs download both on first use and cache them; the Docker image bakes them in and starts fully offline. To pre-install the spaCy model: python -m spacy download en_core_web_sm.
Install Methods¶
Docker is the easiest way to run Lithos — no Python environment management, and both models are baked into the image.
Prerequisites: Docker and Docker Compose (v2+)
This starts Lithos with the HTTP transport on port 8765 — POST /mcp (StreamableHTTP) and GET /sse (legacy SSE) on the same port.
Data is stored in ./data next to the compose file by default; set LITHOS_DATA_PATH to use another directory:
See Docker deployment for multi-environment setups (run.sh, .env.<name> files).
Then start the server:
uv is the fastest way to install Python packages:
Connect an Agent¶
Once Lithos is running, add it to your agent's MCP config. Network clients should prefer the StreamableHTTP endpoint (/mcp); the legacy SSE endpoint (/sse) remains for clients that only speak SSE.
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on your platform:
Restart Claude Desktop. You'll see the Lithos tools in the tool list.
In ~/.openclaw/workspace/config/mcporter.json:
If Lithos is on a different machine:
In Agent Zero's MCP server config (usually in the web UI or mcp_servers.json):
Use host.docker.internal when Agent Zero runs in Docker on the same machine as Lithos.
Verify the Installation¶
You should see 200 OK with:
{
"status": "ok",
"timestamp": "2026-09-05T12:00:00+00:00",
"components": {
"kb_directory": {"status": "ok"},
"search": {"status": "ok"},
"knowledge_base": {"status": "ok"}
}
}
Or from the CLI: lithos stats prints document, index, graph, agent, task, and claim counts plus the data directory in use.
Upgrading¶
Pre-1.0 compatibility
Lithos follows a migration safety over API stability policy pre-1.0. MCP tool signatures may change between minor versions, but your on-disk Markdown knowledge is always preserved. Check the Changelog before upgrading.
Upgrading from 0.3.x or earlier?
v0.4.0 changed the error envelope, and v0.3.2 renamed the sse transport to http (no alias — update any lithos serve --transport sse invocations). See Envelopes, Errors & IDs → Migrating older clients for the full list.
Next Steps¶
- Quickstart → — write your first knowledge item and run a search
- Configuration → — tune data directory, search thresholds, and more