CueAPI makes silent failure impossible.

Schedule agent work, gate every handoff with verification, and refuse to advance until your agent attaches evidence of what it actually did. Once recorded, the outcome is on the record forever — no rewrites, no retroactive edits.

Make your agents accountable.

pip install cueapi
cueapi login
cueapi quickstart

The quickstart command walks you through your first cue end-to-end — schedule a job, run it, report the outcome, see it in the dashboard.

Full docs →

AI agents are unreliable narrators. An agent can return {"success": true} after a tool call that silently failed. It can hallucinate a confirmation. It can report sending an email that never sent.

CueAPI is the primitive that makes the agent prove its claims, at the API boundary, at the moment of reporting. No proof, no progress.

Repo	Description
cueapi/cueapi-core	The open source server. Self-host with Docker.
cueapi/cueapi-python	Official Python SDK. pip install cueapi-sdk
cueapi/cueapi-cli	Official CLI. pip install cueapi (or brew install cueapi/tap/cueapi on Mac)
cueapi/cuechain	TypeScript primitive for verifying contracts between LLM pipeline steps.
cueapi/cueapi-action	GitHub Action to schedule and verify agent jobs from your CI. uses: cueapi/cueapi-action@v1
cueapi/cueapi-mcp	Official MCP server. Give any MCP host (Claude Desktop, Cursor, Zed) scheduler + verification gate tools. npx @cueapi/mcp

• Hosted: cueapi.ai — free tier, no setup
• Self-hosted: cueapi-core — Docker compose, full features, no crippled OSS tier
• Docs: docs.cueapi.ai
• Status: status.cueapi.ai

Made by Vector Apps.