Control AI coding spend. Require proof before merge.

A green CI check is not proof when the same agent can edit the test, workflow, policy, or verifier that defines success.

Runcap adds a required PR check that does not trust the agent or the receipt it hands you. It re-derives the merge decision in a clean CI job, reading the policy and verifier from the pull request's base commit, then replays the permitted change from scratch.

Read the technical explanation: Why a green CI check is not proof.

See real verdicts

Control spend on routed AI calls

Runcap can estimate a cost range and stop routed AI requests once they reach a configured cap. Calls made directly to providers, subscriptions, invoices, cards, or unrelated tools are outside this control path. Multi-agent coding runs can consume far more tokens than a single chat: agents can loop on the same error, rewrite plans, and keep working after the useful change is already complete. For calls routed through the local gateway, Runcap blocks a request before it is sent when that request would exceed the configured cap.

See it in 60 seconds

No API key required. Runcap catches a too-broad request before it spends a cent, then hands back a rescue prompt the moment a run gets stuck.

$ runcap preflight -- claude "build the full app with auth, payments, deploy"

Scope risk: high
Recommendation: Do not launch as one broad mission.
  Split into one vertical slice with a verification command.

$ runcap run --label demo -- npm run build

Error [ERR_MODULE_NOT_FOUND]: Cannot find package '@/components'

Runcap mission: 20260601T221531-demo
Status: stuck (medium confidence)
Changed files: 0   Parsed errors: 1
Primary recommendation: Resolve missing import before continuing

Rescue prompt: Do not continue broad implementation. Diagnose this
  missing module first: Cannot find package '@/components'. Check
  package.json, tsconfig paths, and the latest git diff. Make the
  smallest change that resolves the import, then run the command again.

Trim re-read tokens with delta encoding

Coding agents read a file, change one line, then re-read it. The two copies are almost identical. Runcap can send a lossless line-diff against the version the model already saw, and the model reconstructs the current file from it. Same answer, fewer prompt tokens on that call.

Who this is for

Built for technical founders and engineering teams using Claude Code, Codex, Cursor, or custom coding agents with GitHub pull-request workflows. Runcap is a developer tool: a local gateway for spend control, and a pinned GitHub Action for the Proof Gate.

Two ways to install

Start with local spend control, or wire the Proof Gate into a GitHub repository. They are independent: you can run either on its own.

# run the demo with zero setup
git clone https://github.com/kirder24-code/ai-agent-manager.git
cd ai-agent-manager
npm run setup
npm run demo

# or install the CLI globally
npm install -g runcap

# put a hard daily ceiling on routed Claude or OpenAI calls
ANTHROPIC_API_KEY=sk-ant-... AIM_DAILY_BUDGET_USD=5 runcap gateway
# routed calls are stopped once spend crosses your ceiling

Make the Proof Gate a required PR check in your own GitHub + Node/npm repo. The adjudicator runs from a pinned, immutable Runcap release commit that lives outside the candidate PR, so a pull request cannot rewrite its own judge.

1. Add .runcap/mission.yaml with your scope, protected paths, and verification command.
2. Copy examples/runcap-adjudicate.yml into .github/workflows/ (pinned to the Runcap v0.6.0 release commit).
3. Configure the hardened GitHub branch profile (protected branch, required check, CODEOWNERS for workflow/policy/verifier/protected paths).
4. Make Runcap adjudicate a required status check, so it must pass before merge.

Available today