@getbourdon/client

@getbourdon/client

The Bourdon L6 federation MCP client for TypeScript/Node. Query a running Bourdon L6 server — over a spawned python -m bourdon.core.l6_server stdio subprocess, or a long-lived Streamable-HTTP service — from any JS agent.

Wire-compatible with the Python L6 server: results are JSON-parsed from the server's TextContent envelope, exactly as a Python peer does. Built on @modelcontextprotocol/sdk.

import { BourdonL6Client } from "@getbourdon/client";

// Spawn the Python L6 server over stdio (Python is the oracle):
const bourdon = new BourdonL6Client({ transport: "stdio", library: "~/agent-library" });
await bourdon.findEntity({ name: "Fabrikam" });
await bourdon.listRecentWork({ limit: 10 });
await bourdon.prepareRecognitionContext({ prompt: userMessage });
await bourdon.close();

// Or a remote federation peer over Streamable HTTP with a Bearer token:
const peer = new BourdonL6Client({ transport: "http", url: "https://peer/mcp", token });
await peer.listAgents();

All twelve L6 tools are exposed: queryAgentMemory, listRecentWork, findEntity, listWorkstreams, listNotes, listAgents, exportAgents, getCrossAgentSummary, prepareRecognitionContext, getDeeperContext, compileCodexTurn, commitToFederation. The server applies its own per-tool access_level defaults (public for queries, team for recognition), so omitted args are dropped rather than defaulted client-side.

License: Apache-2.0 (interop client). The Bourdon engine is BUSL-1.1; the canonical implementation is the Python project.