Claude Code, meet the Firetiger plugin! Giving Claude Code super powers in prod
Firetiger makes software in production should be self-operating and self-improving. Our agents understand your observability data, codebase, and the outcomes you care about, and make sure they come true. Think of our agents as little tooth fairies that keep the "bad news" monsters away and leave treats under your pillows, in our case software improvements.
And as of today, you can summon the help of those Firetiger agents directly from Claude Code!
Firetiger is now in the Claude Marketplace.
/plugin install firetiger@claude-plugins-officialHow do you give superpowers to a coding agent?
Today, you can install Firetiger directly from the Claude Code marketplace, a curated set of plugins that give agents a combination of MCP servers and skills. With the Firetiger plugin, Claude Code gets eyes on your production environment and can monitor, debug, and fix issues without ever leaving your terminal.
What actually ships in the plugin
We've included six skills into this plugin, each one designed so Claude can collaborate with our agents to do real work instead of just talking about it:
/firetiger:setup : Creates your Firetiger account, connects your integrations like Vercel, Slack, GitHub, or Linear, and gets your monitoring agent up and running.
/firetiger:instrument : Detects your stack, grabs your OTLP credentials, and drops in the minimal OpenTelemetry instrumentation to start sending traces. Already have OTel? It'll update your exporter config instead. Firetiger setup for those who have an account with us.
/firetiger:query : SQL against your production telemetry. Traces, logs, metrics, all queryable from your terminal. Find the slow spans. Calculate p99 latency. No dashboard required.
/firetiger:investigate : Opens a time-bounded investigation session. You describe the symptoms, Claude digs through the telemetry, and you converge on a root cause together.
/firetiger:create-agent : Describe what you want monitored in plain English. "Watch the checkout API for error rate spikes." Firetiger's planner agent works with Claude to nail down the details, and if it needs input from you, Claude asks. Then your agent is live, watching, and ready to page you (or better yet,
fix it).
/firetiger:monitor-deploy — Point it at a PR. After merge, it checks in at 10m, 30m, 1h, 2h, 24h, and 72h post-deploy. Like a diligent coworker who actually follows up.
Almost anything you can do on the Firetiger UI, you can now do directly from Claude Code!
The loop that closes itself
You write code and ship it. Then a customer complains that something seems broken. Was it related to the release from a day ago? ?!?$% You opening seventeen tabs, squinting at dashboards, muttering things you can't put in a commit message, and roll back your release because you don't have any better ideas.
Here's what happens now instead:
- Your Claude Code instance produces a change and works with Firetiger to figure out how to validate whether it's working well in prod once it is deployed
- Post deploy, a Firetiger agent spots something wrong in production
- It catalogs the issue and collects full context: the impact, blast radius, suspected root cause, etc
- You open Claude Code and mention the issue ID
- Claude pulls the report, writes the fix, and builds a new release that actually works.
All autonomously, your software is more reliable and better driving toward outcomes.
The number of times one of our customers has said "wait, it just... figured it out?" has not gotten old.
Installation
Run this in Claude Code to get the goods!
/plugin install firetiger@claude-plugins-officialThe first run of the plugin will trigger an OAuth flow to sign in to Firetiger, and then you're good to go. Everything runs through our MCP server, the same API powering the Firetiger UI, our Cursor plugin, and our own agents.
If you're ready to stop operating your own software, run /firetiger:setup and it'll get you sorted, or sign up at firetiger.com if you prefer.