Agent skill

The Flare agent skill gives AI coding agents the knowledge to use the Flare CLI to interact with the Flare API. When installed, agents like Claude Code, Cursor, and Codex can access your error data, project data, team data, and user data to triage errors, investigate stack traces, resolve issues, and manage projects.

What are agent skills?

Agent skills are portable instruction files that teach AI coding agents how to use specific tools and APIs. Unlike MCP servers, which require per-client server configuration, skills are simple files that live in your project and work with any agent that supports them.

Installation

First, make sure the Flare CLI is installed and you are logged in:

composer global require spatie/flare-cli
flare login

Then add the Flare skill to your project:

npx skills add spatie/flare-cli

This adds a skill file to your project that any compatible AI coding agent will automatically pick up.

What can it do?

Once installed, your AI coding agent can:

• Triage errors — List open errors, prioritize by occurrence count, and resolve or snooze them in batches.
• Investigate occurrences — Fetch full stack traces with application frames, request context, event trails, and AI-generated solutions.
• Debug with local code — Cross-reference stack trace frames with your local codebase to pinpoint the source of a bug.
• Manage projects — Create new projects, retrieve API keys, and check error counts.
• Manage teams — View team members and their roles.

For example, you can ask your agent things like:

• "Show me the latest open errors in my Flare project"
• "Investigate the most recent RuntimeException and suggest a fix"
• "Resolve all the errors related to the migration we just ran"
• "Create a new Flare project for our staging environment"

When to use this vs MCP

Flare also offers an MCP server that gives your AI agent access to the same data and actions. The agent skill is generally the easier option — a single npx skills add command and you're done — while the MCP server requires per-client configuration but provides a direct connection without the CLI.

The AI development landscape is evolving quickly, so it's worth checking what your preferred agent supports best.