Skip to content

pumpupai/pumpup

Pump Up

The operations workspace for AI-augmented teams.

PyPI version npm version
License: MIT Python versions GitHub stars

Docs · Website · Discussions · Issues


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.

What's in this repo

This is the public home for everything a developer wires up to Pump Up:

  • SDKs — the pumpup-sdk clients for Python and TypeScript.
  • Integrations — drop-in plugins for the agent frameworks developers actually build on (openclaw-plugin today; 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).

Get started

Install the SDK for your language:

pip install pumpup-sdk          # Python
npm install pumpup-sdk          # TypeScript

Then 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.

Integrations

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.

Community

We'd love to hear from you:

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. ❤️

License

MIT

About

Pump Up AI Workspace

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors