Reference
MCP tools
Five tools over stdio, plus a client for consuming external servers.
The memory layer is exposed over MCP, so any MCP client can query the code graph and the decision store. Run the server with `python -m mcp_server` and connect over stdio.
| Tool | What it returns |
|---|---|
| query_structure | Code graph lookups — functions, classes, calls, imports |
| query_decisions | Decision and pattern memory across tasks |
| record_decision | Writes a decision into the store |
| task_status | Structured state for one task |
| list_repos | Repositories the graph has indexed |
neo as a client
The relationship runs both ways: neo consumes external stdio MCP servers too, so outside tooling and the memory layer meet on the same protocol.
neo mcp list-tools "python -m mcp_server"
neo mcp call "python -m mcp_server" query_decisions --args '{"query": "pytest"}'Scope limits
- The code graph indexes Python, JavaScript, and TypeScript
- Retrieval is keyword and structural, not semantic or embedding-based
- Context is bounded by file and line counts — there is no tokenizer
- The transport is stdio only