mux
Pool multiple Claude Max accounts. Never hit 429.
The problem
Your agents burn through one Max seat's 5-hour window in hours. You hit 429. You wait. You lose half a day of work, every day. The only fix is more seats — but managing five OAuth tokens, tracking which has headroom, and failing over manually is its own job.
mux pools 2–20 Max accounts behind one endpoint and routes each request to the account with the most headroom. When one hits the wall, the next request flows to a different account. Your agent never sees a 429.
Features
Headroom routing
Every request goes to the account with the most available 5h and 7d budget, calculated per-request.
Automatic failover
429 on one account? Transparent retry on the next. Your agent never sees it. Sub-second handoff.
Token analytics
Per-account utilization, cost estimation, 5h/7d window exhaustion predictions. Know before you blow.
Production-grade billing
We use the same execution model that runs 17 askalf fleet agents in production on a single Max plan.
OpenAI & Anthropic SDK
Any tool that speaks OpenAI or Anthropic API works out of the box. Cursor, Aider, Continue, custom agents.
Audit log + dashboards
Every request logged with account, latency, token count. Per-tenant attribution if you're a team.
Pricing
Starter
2
accounts
$29/mo
You bring the Claude Max accounts. mux handles routing, failover, and analytics. Cancel anytime.
Need more than capacity?
mux solves the capacity problem. askalf solves the platform problem — full agent fleet, dashboards, scheduling, custom tools that actually work, persistent memory. If you're running serious autonomous agents, you probably want both.
Get early access
Manual onboarding while we ramp. We'll set you up personally and walk you through it.