Estimate before you build
Runcap prices the run up front as a range plus a hard cap, with a truth label on every number. No exact-oracle promises, agent runs are stochastic.
Free, MIT, 100% local
Your AI coding agent is quietly burning your budget.
Runcap is a local-first command-line tool for AI coding workflows. It estimates the cost before you build, hard-caps the spend so the run physically stops at your ceiling, losslessly compresses every call, and flags the agent when it starts circling the same failure. Your code and tokens never touch a server.
Click to copy. Free forever, no signup.
37.9%
tokens saved on a real OpenAI call (1,186 to 737), lossless
$0
the spend after a hard cap fires, the run stops at your ceiling
3 prompts
of 92%+ similarity with no progress trips the loop guard
Hard cap that actually stops
Set a ceiling and the run physically stops the second it hits the cap, instead of discovering the overspend on the invoice.
Loop detection
When 3 prompts in a row are 92%+ similar with no progress, Runcap flags the loop before it burns more budget.
In development
The CLI is free. Cloud is coming for teams.
The local CLI stays free and open source forever. Runcap Cloud is the hosted layer for teams who need shared budgets and alerts. It is being built now. Join the waitlist to get early access and locked-in founder pricing.
- Shared team budgets across everyone's agents
- Alerts on Slack, email, or Telegram when a cap breaks or a run gets stuck
- A dashboard of spend per project and per client
- Founder pricing locked in for everyone on the waitlist