Integrations¶

Connect LLM Council to your existing tools and workflows.

Available Integrations¶

Integration	Description	Use Cases
n8n	Workflow automation platform	Code review, ticket triage, content validation

Integration Patterns¶

LLM Council supports multiple integration patterns:

HTTP API (Recommended for Automation)¶

Use the HTTP API for workflow automation tools like n8n, Make, or Zapier:

# Start the HTTP server
llm-council serve --port 8000

# Call the council endpoint
curl -X POST http://localhost:8000/v1/council/run \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Review this code for security issues: ..."}'

MCP Server (For AI Assistants)¶

Use the MCP server for AI assistants like Claude Code or Claude Desktop:

claude mcp add llm-council --scope user -- llm-council

Configuring MCP Client Timeouts¶

The council's high and reasoning tiers can take 3–10 minutes when frontier reasoning models review large inputs (e.g. multi-thousand-line ADRs). MCP clients have their own transport-layer timeout that is independent of the server's tier budget. If the client times out first, you'll see "MCP layer timeout" errors even though the council is still working — see ADR-012 and issue #327 for the underlying constraint.

Set MCP_TIMEOUT (milliseconds) on the client to at least the tier's server budget:

Tier	Server budget	Minimum client `MCP_TIMEOUT`
`quick`	~30s	`60000` (1 min)
`balanced`	~90s	`120000` (2 min)
`high`	~180s	`300000` (5 min)
`reasoning`	~600s	`900000` (15 min)

In Claude Code, set it per-server in .mcp.json:

{
  "mcpServers": {
    "llm-council": {
      "type": "stdio",
      "command": "llm-council",
      "args": [],
      "env": {
        "MCP_TIMEOUT": "900000"
      }
    }
  }
}

Or set it globally as a shell env var before launching the client:

export MCP_TIMEOUT=900000

This is the only fix for "consult_council times out at high tier" — the server is running fine, the client is hanging up early.

Python SDK (For Custom Applications)¶

Use the Python SDK for custom integrations:

from llm_council import consult_council

result = await consult_council("Should we approve this PR?", verdict_type="binary")

Webhook Callbacks¶

LLM Council supports webhook callbacks for async notifications:

HMAC-SHA256 signatures for request verification
Configurable events (stage completion, errors)
Retry with exponential backoff

See n8n Integration for webhook configuration examples.

Coming Soon¶

Zapier integration
Make (Integromat) templates
Slack bot integration
GitHub Actions workflow