One of the most common questions since we open-sourced DevOps Open Agent:
MCP is great, but how do you stop users from pointing it at any random MCP server?
Fair question. And a good one.
When you connect MCP, your AI can call real tools on real systems: repos, issues, incidents, workflows. That’s powerful for DevOps. It also means an unrestricted paste-any-URL setup isn’t something most teams want in production.
We listened. The latest MCP update in DevOps Open Agent adds layered URL controls:
✅ Platform allowlist: Admins can limit all users to approved MCP endpoints (MCP_ALLOWED_SERVER_URLS)
✅ User whitelist: Name and save trusted servers; once enabled, only those URLs can be used
✅ User blacklist: Block specific MCP servers, even if they’d otherwise be allowed
✅ Official server catalog: Start from supported remote MCP servers (GitHub, Linear, Sentry, etc.) instead of unknown URLs
Checks run when you save settings, test a connection, ask MCP, and during AI investigation enrichment, not just on day one.
This is what open source should feel like: ship → get honest feedback → tighten the product. Thank you to everyone who tried DevOps Open Agent and pushed us on security. Keep the feedback coming.
Configure under Integrations → MCP:
🔗 github.com/ideaweaver-ai/devops-open-agent
🎥 Demo: https://lnkd.in/eBWcPRtj
We’re also going deeper at the intersection of DevOps and GenAI. If that’s your world, check the links
Configure under Integrations → MCP:
🔗 github.com/ideaweaver-ai/devops-open-agent
🎥 Demo: https://lnkd.in/eBWcPRtj
🎓 Want Live Instructor-Led Training?
If you'd like hands-on, instructor-led sessions where you can ask questions and get your doubts resolved live, check out our courses:
Top comments (0)