About conductor-graph-mcp
conductor-graph-mcp presents a CognOS agent system as a machine-readable JSON graph for a conductor such as Claude. It provides a system snapshot and queries for node status, communication edges, and blocked nodes.
Key features
- 01Retrieve the full agent graph
- 02Check a node's live status
- 03View communication edges
- 04List blocked nodes
Use cases
- 01Review the system structure
- 02Find currently blocked or failing nodes
- 03Inspect communication between nodes
MCP support
The MCP server exposes graph-query tools to MCP clients, including a full snapshot, node status, edges, and blocked-node information.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
uvx conductor-graph-mcpCOGNOS_BASE_URLURL of the TrustPlane/CognOS gateway (default: http://127.0.0.1:8788)COGNOS_API_KEYSecretAPI key for the CognOS gateway (optional, leave empty if not required)Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEConductor Graph MCP Server
mcp-name: io.github.base76-research-lab/conductor-graph-mcp

An MCP server that exposes a CognOS agent system as a machine-readable JSON graph. The conductor (Claude) calls a single tool and gets an immediate full system snapshot.
Built on the FNC architecture: Field (agents) → Node (processing) → Cockpit (oversight)
Installation
# From PyPI (recommended)
pip install conductor-graph-mcp
# Or from source
git clone https://github.com/base76-research-lab/conductor-graph-mcp
cd conductor-graph-mcp
pip install -e .
Usage with Claude Code
Add to ~/.claude/settings.json under mcpServers:
{
"mcpServers": {
"cognos-graph": {
"command": "python3",
"args": ["-m", "conductor_graph_mcp"],
"env": {
"COGNOS_BASE_URL": "http://127.0.0.1:8788",
"COGNOS_API_KEY": "your-key"
}
}
}
}
Or with uvx:
{
"mcpServers": {
"cognos-graph": {
"command": "uvx",
"args": ["conductor-graph-mcp"],
"env": {
"COGNOS_BASE_URL": "http://127.0.0.1:8788",
"COGNOS_API_KEY": "your-key"
}
}
}
}
Restart Claude Code — the server activates automatically.
Note: Requires MCP SDK ≥ 1.0 (
mcpon PyPI).
Tools
| Tool | Description |
|---|---|
get_agent_graph | Full graph — nodes + edges + summary |
get_node_status(node_id) | Live status for a specific node |
get_edges | Only edges (communication flow) |
get_blocked_nodes | Shortcut: what is broken right now? |
Known Node IDs
conductor
trustplane-gateway
token-compressor
session-memory
armada-bus
agent-critic
agent-curator
agent-ethics
agent-synth
agent-self
agent-srt
agent-indexer
Examples
get_agent_graph() # Full system overview
get_node_status("trustplane-gateway") # One specific node
get_blocked_nodes() # What is broken?
get_edges() # Communication flow
Node Status — Data Sources
| Node | Source |
|---|---|
trustplane-gateway | HTTP GET /healthz + /v1/providers/health |
token-compressor | /tmp/b76_compress.log |
session-memory | /tmp/b76_save.log + SQLite traces.sqlite3 |
armada-bus | /tmp/b76_armada_bus.json + /tmp/b76_armada_bus.log |
agent-* | psutil — is the process running? |
conductor | Always active (the conductor itself) |
Environment Variables
| Variable | Default | Description |
|---|---|---|
COGNOS_BASE_URL | http://127.0.0.1:8788 | TrustPlane gateway URL |
COGNOS_API_KEY | (empty) | API key for gateway auth |
Verification
# Test that the server starts
python3 -m conductor_graph_mcp
# Stop the gateway → trustplane-gateway.status = "error"
# get_blocked_nodes() should return trustplane-gateway
# Run a /save → session-memory.metrics.entries increases
# get_node_status("session-memory") shows new count
License
MIT — Base76 Research Lab
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.