Reflex is Live

I shipped Reflex today.

It's error tracking. But not like you've seen before.

The moment that sparked this

I was debugging a production issue last month. The usual dance: open Sentry, find the error, look at the stack trace, check the logs, try to reproduce, scratch my head, repeat.

Midway through, I asked Claude for help. Pasted the stack trace. Claude analyzed it, made a guess about the root cause.

"Can you show me the logs from around that time?"

I sighed. Opened another tab. Found the logs. Copied them. Pasted them back.

"And the database state for that user?"

Another tab. Another query. Another copy-paste.

Fifteen minutes of playing messenger between Claude and my tools. And I thought: this is so stupid.

Claude can analyze errors brilliantly. But it can't see them. I'm the bottleneck.

That's why Reflex exists.

What makes it different

Reflex gives Claude direct access to your errors via MCP.

Ask: "Why is checkout failing?"

Claude queries Reflex. Finds the errors. Analyzes the stack traces. Cross-references with Recall (logs). Comes back with:

"The checkout is failing because stripe_customer_id is nil for users created before December 15th. The backfill migration didn't run in production. 47 users affected. Here's the fix..."

Not just the error. The root cause. The impact. The solution.

No copy-pasting. No tab-switching. Just ask and get answers.

Smart grouping

This is the part I'm proudest of.

Traditional error tracking groups by exception class and stack trace. So you get:

NoMethodError: undefined method 'name' for nil:NilClass
  at app/models/user.rb:42

NoMethodError: undefined method 'email' for nil:NilClass
  at app/models/user.rb:67

Sentry sees two different errors. But they're the same problem: something's returning nil when it shouldn't.

Reflex groups by semantic meaning. Both errors become one issue: "User model returning nil unexpectedly."

Smarter grouping = less noise = faster resolution.

The integration is stupid simple

# Gemfile
gem 'brainzlab'

# That's it.

Reflex hooks into Rails automatically. Exceptions get captured. No manual begin/rescue wrappers. Context gets attached—current user, request params, environment, everything.

Zero-config error tracking. As it should be.

What else it does

Correlation with Recall. Every error links to the logs that happened around it. "Show me what happened in the 5 minutes before this error" just works.

Suggested fixes. Reflex learns from errors it's seen. When a similar pattern appears, it suggests solutions.

Resolution tracking. Mark an error as resolved. Reflex watches for regressions. If it comes back after a deploy, you'll know immediately.

MCP-first design. Everything is queryable. Natural language in, structured data out. Built for AI from day one.

Try it

Five minutes to get it running:

git clone https://github.com/brainz-lab/stack.git
cd stack
./scripts/setup.sh
docker-compose up -d reflex timescaledb redis traefik

Open http://reflex.localhost. You're in.

The first of many

Reflex is product #1 of 31.

It's not perfect. There are rough edges. The UI needs work. Some features are missing.

But it works. It shipped. Real code, running in the real world.

That's what matters right now. Ship, learn, improve, repeat.

Onward to product #2.

— Andres