In our previous post we introduced the Flare CLI. Today we want to highlight one of its most powerful features: the AI agent skill.
With a single command, you can give AI coding agents like Claude Code, Cursor, or Codex direct access to your Flare data:
flare install-skill
That's it. No MCP server to configure, no separate processes to manage. The skill file gets added to your project and your agent picks it up automatically.
Once installed, your agent can look up errors, analyze stack traces against your local code, triage and resolve issues, and review performance data for your routes, jobs, and queries, all without you leaving the terminal or copy-pasting from a dashboard.
It works the other way too. Ask your agent to review the performance of your app, and it'll pull in the data from Flare and give you actionable suggestions.
Install the Flare CLI, run flare install-skill in your project, and see what your agent can do with it.
Continue reading
Why we removed AI solutions from Flare
AI generated solutions are dead, long live AI generated solutions! We removed the old one-shot fixes from a model that could only see the stack trace. It simply couldn't compete with modern agentic AI. In their place, Flare works with the agentic AI tools you already use: an MCP server that handles real triage, a CLI with an agent skill, quick Copy for AI prompts in the UI and docs served as markdown. Hosted agents are in the works too.
Alex
Flare now uses OAuth for MCP, the CLI, and API
Connect MCP clients like Claude and Cursor to Flare by logging in from your browser, no more copying API tokens. Every connection gets its own permissions, so you can give your AI read-only access to a single project. The CLI signs in the same way, and new personal access tokens with scopes and expiry cover scripts and CI.
Alex
Subscribe to Backtrace, our quarterly Flare newsletter
No spam, just news & product updates
