When you ship an agent into production, it handles most of the work — and then a human still has to approve, route, or decide on the slice it can't. Pump Up is where that handoff lives: the place agents drop work for humans, and the place humans approve / route / decide.
It's the control panel for the team that processes the residual 30% of agent work — reviewing, approving, routing, handling exceptions. One coherent surface: queue → item view → audit log → manager view. The agent declares the work; we render it beautifully and keep an immutable record of who decided what and why.
We're agent-agnostic — run agents from any vendor (or your own) and Pump Up ingests work from all of them.
This is the public home for everything a developer wires up to Pump Up:
- SDKs — the
pumpup-sdkclients for Python and TypeScript. - Integrations — drop-in plugins for the agent frameworks developers actually build on (
openclaw-plugintoday; more driven by where you are). - Docs (soon) — guides and the full API reference at docs.pumpup.com.
Note
This repo is a read-only mirror. The code is authored in an upstream monorepo and synced here automatically — so the history you see is squashed sync commits, not the real development log. See CONTRIBUTING.md for how to get involved (spoiler: Issues and Discussions, not PRs).
Install the SDK for your language:
pip install pumpup-sdk # Python
npm install pumpup-sdk # TypeScriptThen drop a human in the loop in about 20 lines — a single call parks the work, routes it to the right person, and hands you back the decision:
const decision = await pumpup.requestApproval({
type: "refund_over_threshold",
context: { customer, amount, agentRecommendation },
schema, // the agent declares the shape; we render the item view
sla: "4h",
routingHint: { team: "claims" },
idempotencyKey, // exactly-once, safe to retry
});The human reviews it in a fast, keyboard-driven item view; your agent gets the answer back. Full walkthrough at docs.pumpup.com.
Already building on a framework? Skip the wiring:
- OpenClaw — request approvals/elicitations and resume your agent when the decision lands.
More integrations land here as siblings — open a Discussion to tell us which framework you want next.
We'd love to hear from you:
- 🐛 Found a bug or hit a rough edge? Open an issue.
- 💡 Want a feature or a new integration? Start a discussion.
- 🔒 Found a security issue? See SECURITY.md — please report privately.
One thing up front: because the code is authored upstream and mirrored here, pull requests are turned off on this repo — there's nowhere for them to land. Issues and Discussions are where the real conversation happens, and we're listening. ❤️