MCP Server

Status: in development, not yet available

A native MCP server for Shift-Left Studio is in development. It is not shipped today — there is no backend/src/mcp/server.js process to run, no tenant configuration to set, and no tier includes it yet. This article describes what is planned, not a feature you can set up right now.

If you're looking for a way to drive test generation from an AI agent today, use the Public API instead — see below.

What's planned

The goal is to let external AI agents drive Shift-Left Studio's test-generation pipeline through the Model Context Protocol (MCP), so any MCP-capable client — Claude, Cursor, and others — could call tools to analyze an endpoint, generate and validate tests, and explain coverage, and read live coverage/gap/test/schema data as resources. The design intent is a stdio-based server, scoped per tenant, so specs and prompts stay inside your environment.

None of this is available today. We don't have a public ship date.

Use the Public API today

Every action the planned MCP tools would perform is already available through the Public API: authenticate, trigger test generation, poll status, and fetch results. Any agent framework capable of making HTTP calls — including custom tool definitions in Claude, Cursor, or other agent runtimes — can be wired to it now, with more setup than a native MCP connection will eventually require, but it works today.

Related articles

• Generate endpoint test cases
• Coverage and fidelity
• Public API
• AI Settings