Hi Cursor community,
We all love how powerful Cursor’s Composer and Agents are, but let’s be honest — sometimes agents can be a bit too confident. They might rush into coding without reading your project guidelines, push directly to the main branch, or accidentally ignore your existing CI/CD standards.
To solve this, I built agentic-sdlc-mcp, an engineering governance and safety control plane for AI coding agents. It acts as a protective shield between Cursor and your GitHub repositories.
Key Features:
- Strict Context Alignment: Forces agents to plan based on your exact issue context,
CLAUDE.md, or architecture guidelines before writing code. - Safety First (
dryRun: trueby default): All destructive or mutating tools (like creating PRs/Issues) run in dry-run mode first to let you review. - Quality & Security Gates: Agents can automatically audit workflow permissions, check branch protection status, and run security triages before making changes.
- 12 Dedicated Tools: Built-in tools for
plan_from_context,review_pr_against_standard,prepare_work_item, and more.
How to use it in Cursor
It takes less than 30 seconds to configure. Just open your Cursor settings, go to Models -> MCP, and add a new MCP server with:
- Type:
command - Command:
npx agentic-sdlc-mcp configure(or run it in your terminal to save global config)
Check out the full open-source project and architecture diagram here:
GitHub: GitHub - SakuraCianna/agentic-sdlc-mcp · GitHub
I’d love to hear your thoughts! How do you currently enforce engineering discipline when letting Agents roam free in your codebases?