cashpilot-mcp is an MCP (Model Context Protocol) server whose role is to act as a bridge layer, exposing any CashPilot instance to LLMs. CashPilot itself is a passive income cluster manager, and cashpilot-mcp lets large language models monitor passive income earnings, manage services, and control cluster worker nodes.
In terms of purpose, it is aimed at users who use CashPilot to manage passive income services such as bandwidth sharing, and who want to view earnings and operate services conversationally through an LLM or MCP client.
Its main capabilities include:
- Read-only resources: providing read-only resources such as earnings, deployed services, service catalog, cluster status, worker nodes, health scores, and collector alerts, for example cashpilot://earnings/summary, cashpilot://fleet/summary, etc.
- Tools: supporting querying daily and historical earnings, reading service logs, and starting, stopping, restarting, deploying, or removing services.
- trigger_collection: run an earnings collection once immediately across all services.
- get_compose: return the Docker Compose definition of a service.
In terms of usage, it communicates with the user's own CashPilot instance through an admin API key, and the key is CashPilot's admin API key (not the cluster key). Communication supports HTTP or stdio: when using HTTP, it goes through a single JSON-RPC endpoint /mcp; when using stdio, set TRANSPORT=stdio. By default it listens on 127.0.0.1:8081; when it needs to be exposed externally, MCP_AUTH_TOKEN can be used to add bearer authentication. It provides a Docker image, an npm package (npx cashpilot-mcp), and multi-architecture Go binaries. Register the npm package in an MCP client (such as Claude Desktop, Claude Code, Cursor), and npx will run it in stdio mode.
The applicable scenarios are mainly users who want to integrate CashPilot's earnings monitoring and service management into LLM workflows, for example viewing earnings and managing them in natural language through clients such as Claude or Cursorservices.
In terms of limitations, the project uses the GPL-3.0-or-later license. Details regarding the specific supported operating systems, dependency versions, performance, pricing model, etc. are not yet provided in the materials.
