mcpdirectory.dev
9,129 servers · updated 23 Sept

Mendapi MCP server

official from Mendapi Active last commit 49 d
{
  "mcpServers": {
    "mendapi": {
      "command": "npx",
      "args": [
        "-y",
        "mendapi@latest"
      ],
      "env": {
        "OPENAI_API_KEY": "sk-...",
        "STRIPE_KEY": "sk_live_..."
      }
    }
  }
}
file .mcp.json (in your project root)install npx -y mendapiversion 0.5.6

What it does

Mendapi detects breaking changes in upstream APIs you depend on, scans your codebase to find what's actually affected, and drafts migration fixes. It runs entirely locally and is designed for teams that need their code to never leave their machines. This server exposes mendapi's scanning and fixing capabilities to MCP clients like Claude Code and Cursor.

  • Scan a repository against a local database of breaking changes from 20 major API providers to identify exactly which files and symbols are affected.
  • Generate a reviewable diff that migrates your code away from deprecated API calls using deterministic rule packs for known migrations like OpenAI v3→v4 or Stripe v7→v8.
  • Integrate API change detection into your CI pipeline to catch breaking changes before they break your integration, with results stored and audited locally.

Tools this server exposes

ToolDoes
syncFetch the upstream API change feed from mendapi infrastructure into a local SQLite database; the only command that touches the network.
scanRead the local change database and detect which breaking changes from upstream providers actually hit your codebase, reporting file, line, symbol, and confidenc
fixApply migration rule packs to generate a reviewable diff that migrates your code away from breaking API changes, without modifying files until you pass --apply.
reviewPerform an optional LLM semantic review pass on scan results to filter low-confidence findings.
prCreate a fix branch, commit with migration guide citations, and prepare a PR description based on scan results.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.