Leanmcp/gateway-skills

What it solves

Leanmcp provides visibility and control over AI coding agents. It solves the "black box" problem where developers cannot see exactly what prompts, system instructions, or tool calls their agent is sending to the model provider, as well as the lack of transparency regarding which files are being read into the context and the associated token costs.

How it works

It operates as a drop-in proxy (AI Gateway) that sits between the AI coding agent (such as Claude Code) and the model provider. By changing a single environment variable for the base URL, all traffic is routed through the gateway, which logs requests and responses, tracks file visibility, and monitors token usage and costs via a dashboard.

Who it’s for

Developers using AI coding agents who need to audit agent behavior, track costs per project or session, and ensure transparency in how their codebase is being accessed by the AI.

Highlights

• Zero-code integration: Works by updating one environment variable without changing agent code.
• Full observability: Provides logs for exact prompts, system prompts, and tool calls.
• Context tracking: Visualizes which files the agent read and pulled into context.
• Cost monitoring: Tracks tokens and costs on a per-request, per-session, and per-project basis.
• Skill library: Includes a collection of "skills" for MCP development, inference, training, and research workflows.