Fastn.ai MCP Server — Unified API Integrations

Overview

Fastn.ai MCP Server provides a single, unified API layer to connect applications to a broad ecosystem of tools, actions, and workflows. Instead of integrating dozens (or hundreds) of service-specific SDKs and webhooks, you talk to one server that proxies, normalizes, and secures access to remote systems. The server includes built-in authentication, real-time monitoring, and a catalog-style interface for discovering and invoking capabilities.

This becomes useful when you need consistent tooling across analytics pipelines, internal apps, or AI-driven workflows. Developers can programmatically list available tools, call actions, orchestrate multi-step workflows, and stream events back to clients without implementing separate integrations for each backend service.

Features

• Single unified API for tools, actions, and workflows
• Built-in authentication and role-based access control
• Real-time event streaming and monitoring (events, logs, metrics)
• Catalog of 1,000+ connectors (databases, storage, messaging, ML, SaaS)
• Action invocation with structured inputs/outputs and async handling
• Workflow composition and chaining of actions
• Pluggable adapters for custom connectors and enterprise systems
• HTTP(S) API and WebSocket/Server-Sent Events (SSE) for live updates
• Audit logging and request tracing for observability

Installation / Configuration

Clone the repository and run with Docker Compose (recommended for quick local setup):

git clone https://github.com/fastnai/mcp-fastn.git
cd mcp-fastn
docker-compose up --build

Configuration is driven by environment variables. Example .env:

MCP_HOST=0.0.0.0
MCP_PORT=8080
MCP_API_KEY=your_secret_key_here
DATABASE_URL=postgres://user:pass@db:5432/mcp
REDIS_URL=redis://redis:6379/0
LOG_LEVEL=info
ENABLE_METRICS=true

You can also run the server without Docker by installing dependencies and launching directly (example uses Python-style invocation if applicable):

# install dependencies (example)
pip install -r requirements.txt

# run
MCP_API_KEY=your_secret_key python -m mcp_fastn.server --host 0.0.0.0 --port 8080

Security notes:

• Protect MCP_API_KEY and rotate keys regularly.
• Place the MCP server behind your internal network/WAF for production.
• Integrate with your identity provider or proxy for RBAC if required.

Available Resources

The MCP server exposes a catalog of resources organized by category. Typical categories include:

You can query the catalog with standard endpoints (examples below) to list tools, inspect inputs/outputs, and view required credentials.

Quick API Examples

List available tools:

curl -H "Authorization: Bearer $MCP_API_KEY" \
  "http://localhost:8080/v1/tools"

Invoke an action synchronously:

curl -X POST "http://localhost:8080/v1/actions/run" \
  -H "Authorization: Bearer $MCP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "tool_id": "postgres.query",
    "input": {
      "connection": "prod-db",
      "query": "SELECT count(*) FROM events WHERE ts > now() - interval ''1 day''"
    }
  }'

Start a long-running workflow and receive events via SSE:

curl -H "Authorization: Bearer $MCP_API_KEY" \
  "http://localhost:8080/v1/workflows/run?workflow_id=etl_pipeline" \
  --no-buffer
# use browser or SSE client to consume live events at /v1/events/stream

Retrieve metrics and logs (if enabled):

curl -H "Authorization: Bearer $MCP_API_KEY" \
  "http://localhost:8080/v1/monitoring/metrics"

Use Cases

• Analytics orchestration: Trigger a nightly ETL workflow that reads from S3, runs transformations on Databricks, and writes to a data warehouse. Use the MCP API to schedule, monitor, and get alerts without per-service SDKs.
• Unified UI for internal tools: Build a single admin dashboard that lists available integrations (Slack, GitHub, databases). The dashboard uses the MCP server for authentication and for executing actions like creating issues or running queries.
• Secure LLM action execution: Allow LLMs to call external actions (e.g., query a database or call an internal service) via a controlled MCP action interface with RBAC and auditing.
• Cross-team automation: Compose multi-step workflows that touch DevOps, analytics, and messaging systems (deploy code, run tests, notify Slack) while keeping sensitive credentials centralized on the MCP server.

Getting Help & Contributing

• Source and issues: https://github.com/fastnai/mcp-fastn
• Open a GitHub issue for bugs or feature requests.
• Contributing typically involves adding adapters to the connectors directory, writing tests, and following the repository’s contribution guidelines.

For production deployments, review the repository README for advanced configuration (high-availability, persistent storage, secrets management) and integrate with your organization’s observability and identity systems.