Badgr Compute API / Build on Badgr

Build on Badgr

The Badgr dashboard is one way to use Badgr. The Badgr Execution API is how other products build on Badgr: use it as the cloud execution layer behind your own AI product — agents, GPU workloads, and model-serving endpoints — without exposing the Badgr dashboard to your users.

They build the app. Badgr runs the work.

Your product, your UI, your users, your branding, your workflows. Badgr owns execution, compute, runtime limits, retries, recovery, verification, artifacts, and teardown underneath it.

Common architecture

1Your product
2Your backend
3Badgr Execution API
4Agent / CPU / GPU / Serve
5Verified result + events + artifacts
6Your product

Your users never need to see the Badgr dashboard — Badgr can operate entirely behind your product.

Choose a use case

One API, many workload types

Every use case above calls the same Jobs API. Here's the Agent Job that a coding-agent product would submit:

POST /v1/jobs

{
  "type": "agent",
  "input": {
    "repository": "https://github.com/acme/payments",
    "instruction": "Fix checkout"
  },
  "policy": {
    "max_cost": 2,
    "max_runtime_minutes": 20
  }
}

Other job types (custom.run, model.serve, train.lora, comfy.batch) run on the same underlying execution platform — see /docs/compute-api for the full job type reference.

The mental model

Your product. Your users. Badgr underneath.

OpenAI is an intelligence API. Stripe is payments infrastructure. Badgr is AI execution infrastructure — the cloud that runs the work your product asks for.