lastsaas

LastSaaS includes a built-in MCP (Model Context Protocol) server that exposes AI-friendly access to a running LastSaaS instance. This MCP server provides a curated set of read-only tools that let an agent query operational data, inspect health metrics, review billing and usage data, examine API keys, and audit system state — all through natural language queries. The tools are designed to be safe and read-only by default, enabling AI agents to monitor and assist with administration without mutating state. You can connect Claude or any MCP-compatible agent to this endpoint to perform tasks like fetching ARR trends, checking recent error spikes, auditing keys, or surveying system health, all from within your existing LastSaaS deployment.

To use the MCP, point your MCP-enabled AI agent at the running LastSaaS MCP endpoint and start with high-signal queries such as: "Show me the current ARR and monthly active users (MAU)." or "List the last 24 hours of system health warnings." The LPC (Low-Privilege Context) design ensures sensitive actions require explicit authorization, while read-only data remains readily accessible for analysis and monitoring.

If you want to customize how the agent interacts with LastSaaS, you can enable or disable specific tools, tweak the data scope, or adjust authentication and rate limits via environment variables and configuration in your deployment. This enables safe, AI-assisted administration without compromising security or data integrity.

Prerequisites:

• Docker installed and running on the host where you deploy LastSaaS MCP access
• Access to a LastSaaS deployment (local or cloud)
• Basic familiarity with containerized deployments

Installation steps (Docker-based):

1. Pull and run the LastSaaS MCP container (example image name used in this repo):

   docker pull jonradoff/lastsaas

2. Run the container with an interactive session and necessary environment variables:

   docker run -it
   --env MCP_ENABLED=true
   --env MCP_AUTH_TOKEN=your-token-or-empty-if-not-required
   jonradoff/lastsaas

3. Verify the MCP endpoint is reachable. If the container exposes a port for MCP, ensure you map it to a host port, e.g. -p 5000:5000, and test with a simple curl or your MCP client:

   curl http://localhost:5000/mcp/health

4. Configure your MCP client or agent to connect to the LastSaaS MCP endpoint and start issuing read-only queries via the provided tools.

Alternative: If you prefer not to use Docker, you can adapt this to your preferred deployment method (Go binary, Fly.io, etc.) once you have the MCP server image or binary available in your environment.

Tips and common considerations:

• The MCP server exposes read-only tooling by default to ensure safe agent-based administration. If you need to perform write actions, verify what tools are enabled and ensure proper authorization.
• When running in production, pin the MCP container image to a specific version to avoid unexpected changes.
• Use environment variables to control access, rate limits, and which tools are exposed to the agent. Typical vars might include MCP_ENABLED, MCP_AUTH_TOKEN, and any tool-specific toggles.
• If you encounter connectivity issues, verify that the MCP endpoint is accessible from the agent's network, and confirm that any required ports are properly published and not blocked by firewalls.
• The MCP interface is designed to be model-agnostic; you can connect Claude, GPT-based agents, or any MCP-compatible interface to inspect ARR, health metrics, API keys, webhooks, and billing status from LastSaaS.