{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibkgb42luhy6lmnyksd6hygjgsxcra272jnhfg4x5xnlsqc4e62r4",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mowitsyoaqq2"
  },
  "path": "/t/coding-tools-mcp-turning-chatgpt-into-a-real-coding-agent/1384178#post_4",
  "publishedAt": "2026-06-23T03:03:16.000Z",
  "site": "https://community.openai.com",
  "textContent": "Nice direction. The thing I would make very explicit in the server docs is the separation between inspection, planning, and mutation tools.\n\nFor a coding MCP, I would want the tool surface to make these boundaries visible before any call runs:\n\n  * read/search/list operations are always safe to call\n  * write/apply/run commands require an explicit working directory and scope\n  * destructive operations require dry-run output first\n  * every command returns enough diagnostic context for the model to recover without guessing\n  * the server logs which files, commands, and git state were touched\n\n\n\nThe MCP part makes the workflow possible, but the trust comes from the contract around side effects.",
  "title": "Coding Tools MCP: Turning ChatGPT into a Real Coding Agent"
}