mcp server
n8n-fabric
Workflow automation fabric layer - n8n MCP server with Qdrant vectors and Redis caching
Set up
Runs on your computer. Pick your agent and operating system.
Before you start
- uv (`curl -LsSf https://astral.sh/uv/install.sh | sh`)
{
"mcpServers": {
"n8n-fabric": {
"command": "uvx",
"args": [
"n8n-fabric"
]
}
}
}same format Cursor, GitHub Copilot CLI, Junie, JetBrains AI Assistant, Augment (Auggie CLI), Factory Droid, Kiro, Devin CLI +12
Values in angle brackets are placeholders. Restart Claude Desktop after editing its config.Claude Desktop docs
README
synced Oct 21, 2026What it does
n8n has 500+ nodes and a full REST API, but it lives behind a UI. n8n-fabric puts it behind MCP instead, so an AI assistant can list, build, run and manage workflows in plain language. Every workflow is also embedded into Qdrant, so you can find one by what it does — "how did I handle webhook → transform → API?" — not just by name. Redis caches metadata and execution state; PostgreSQL is n8n's own store.
The 36 tools
| Group | Tools |
|---|---|
| Workflows | workflow_list · _get · _create · _update · _delete · _activate · _deactivate · _execute |
| Executions | execution_list · _get · _delete · _retry · _stop |
| Credentials | credential_list · _create · _delete · _schema |
| Tags & Variables | tag_list · _create · _delete · variable_list · _get · _create · _delete |
| Projects & Users | project_list · _create · _delete · user_list |
| Community nodes | community_node_list · _install · _uninstall |
| Source control | source_control_pull · _push · _status |
| System | audit_logs · fabric_status |
Quick start
git clone https://github.com/ry-ops/n8n-fabric
cd n8n-fabric
# full stack: n8n + Qdrant + Redis + PostgreSQL
docker compose up -d
docker compose ps
open http://localhost:5678 # n8n UI
Run the MCP server from source (Python 3.11+, uv):
uv sync
export N8N_URL=http://localhost:5678
export N8N_API_KEY=your-api-key
uv run n8n-fabric-mcp
Connect Claude Desktop — add to claude_desktop_config.json:
{
"mcpServers": {
"n8n-fabric": {
"command": "uv",
"args": ["run", "n8n-fabric-mcp"],
"cwd": "/path/to/n8n-fabric"
}
}
}
Then ask: "List my active workflows," "Create a webhook → Slack workflow and activate it," or "Find the workflow that posts deploy notifications and run it."
Configuration
| Variable | Description | Default |
|---|---|---|
N8N_URL | n8n API URL | http://localhost:5678 |
N8N_API_KEY | n8n API key | (required) |
QDRANT_URL | Qdrant URL | http://localhost:6343 |
REDIS_URL | Redis URL | redis://localhost:6389 |
N8N_ENCRYPTION_KEY | n8n encryption key | (required for production) |
Ports
6343/6389are used to avoid clashing with any existing Qdrant/Redis instances.
Part of the ry-ops fabric
n8n-fabric is one layer of a larger fabric, each exposed over MCP:
| Fabric | Role |
|---|---|
| n8n-fabric | workflow execution & playbooks |
| git-steer | repo lifecycle, security, PRs |
| aiana | semantic memory & pattern recall |
Together they close the loop: "Create a webhook-to-Slack workflow and commit it" → n8n-fabric builds it, git-steer commits the JSON, aiana indexes it for next time.
Documentation
- CLAUDE.md — AI assistant context
- Fabric coordination — wiring n8n-fabric, git-steer and aiana together
- ADR-001 — fabric design decisions
License
MIT. See LICENSE.