DEV Community

brian austin
brian austin

Posted on

Claude Code is locking people out — here's how to keep working during outages

Claude Code is locking people out — here's how to keep working during outages

If you're reading this, you've probably been hit by the wave of Claude Code lockouts reported on GitHub this week (issue #44257). Hours locked out. Work stopped. Deadlines missed.

Here's what's happening and — more importantly — how to stay productive regardless.

What's causing the lockouts

The GitHub issue shows a pattern: Claude Code users are hitting aggressive rate limiting that locks their accounts for hours at a time. This isn't a brief pause — it's a multi-hour hard block that completely stops your workflow.

Anthropics own status page shows no incident, but hundreds of developers are reporting the same thing: mid-session, Claude Code stops working and returns an error that doesn't resolve for 2-4 hours.

The suspected causes:

  • Subscription tier enforcement tightening
  • Multi-agent/subagent usage triggering rate limits faster
  • The February 2026 updates changing how token budgets are tracked

The escape hatch: ANTHROPIC_BASE_URL

Here's something most developers don't know: Claude Code supports a ANTHROPIC_BASE_URL environment variable that lets you route requests through a different endpoint.

This was designed for enterprise proxies, but it works for any compliant Claude API endpoint.

export ANTHROPIC_BASE_URL=https://api.simplylouie.com
export ANTHROPIC_API_KEY=your-simplylouie-key
Enter fullscreen mode Exit fullscreen mode

Then run Claude Code normally:

claude
Enter fullscreen mode Exit fullscreen mode

Your requests route through a separate endpoint that:

  • Has its own rate limit pool (not shared with the main Anthropic API)
  • Costs a flat ✌️2/month regardless of how many requests you make
  • Keeps working during Anthropic's lockout periods

Setting up a fallback workflow

The smart approach is to set up Claude Code with two profiles: your normal Anthropic credentials, and a fallback endpoint for when you get locked out.

# ~/.bashrc or ~/.zshrc

# Normal Claude Code
alias claude-normal='ANTHROPIC_API_KEY=your-anthropic-key claude'

# Fallback when locked out
alias claude-fallback='ANTHROPIC_BASE_URL=https://api.simplylouie.com ANTHROPIC_API_KEY=your-simplylouie-key claude'
Enter fullscreen mode Exit fullscreen mode

When you hit a lockout:

# Switch to fallback instantly
claude-fallback
Enter fullscreen mode Exit fullscreen mode

Same Claude model. Same interface. No context switching. You're back to work in 30 seconds.

The deeper problem with subscription AI

The lockout issue exposes a fundamental tension with subscription AI tools: when the provider has capacity issues, your work stops.

The lockout reports on GitHub show developers losing 2-4 hours mid-sprint. For freelancers on client deadlines, that's real money. For team leads, that's blocked engineers.

Having a fallback that costs ✌️2/month (that's literally two dollars) is cheap insurance against this happening to you.

What you can do right now

  1. Check if you're affected: Look for the multi-hour lockout pattern — not just normal rate limiting
  2. Set up the fallback: Takes 2 minutes, costs ✌️2/month at simplylouie.com
  3. Test it works before you need it: Run a quick prompt through the fallback endpoint now so you know it works

The flat-rate advantage during outages

One underappreciated aspect: when you're locked out of the main Claude Code subscription, you might try to "make up time" when you get back in — burning through your token budget fast.

With a flat-rate proxy at ✌️2/month, there's no token budget. Make as many requests as you need during the lockout period. Catch up without worrying about hitting the rate limit again.


Get the fallback set up: simplylouie.com — ✌️2/month, 7-day free trial, no card required.

50% of revenue goes to animal rescue. Building sustainable AI tools for the long haul.

Top comments (0)