MCP (beta)
Use Faraday's MCP server to give agents real-time access to data and predictions.
Faraday's MCP server is a beta feature that allows LLM-powered agents to access your Faraday data and predictions in real time. This enables more dynamic and context-aware interactions with your data, enhancing the capabilities of AI-driven applications.
Faraday setup
- Configure your Faraday resources normally, implementing your preferred use case by creating the necessary datasets, cohorts, objectives, and pipeline.
- Ensure your pipeline is out of preview mode.
- Add a Lookup API deployment (target) to your pipeline. From the deployment's URL, extract the target ID (the UUID after the string /targets/).
- Finally, go to Settings → Account and copy your API key.
Agent setup
Here's how to connect to Faraday's MCP server using MCP Inspector.
In your terminal, run npx @modelcontextprotocol/inspector and copy MCP_PROXY_AUTH_TOKEN from the logs in the terminal.
Then, open MCP Inspector UI at http://localhost:5173 and:
- Set URL to https://mcp-beta.faraday.ai/mcp
- Expand "Authorization" section and set "Header Name" to Authorization and "Bearer Token" to the API key you saved above
- Expand "Configuration" section and set "Proxy Session Token" to the MCP_PROXY_AUTH_TOKEN you saved above
- Click "Connect"
- Under "Tools" click "List Tools" and then faraday-lookup-tool.
- In the resulting interface, you will input the target_id you saved above, along with the identifiers of the person you want to retrieve context on.
Getting help
This is a beta feature, so some aspects may change. Please don't hesitate to reach out to your account team or open a ticket in our support portal if you have any questions or feedback.