For developers

The workload catalog,
one import away.

A Python SDK, a TypeScript SDK, a CLI, and an MCP server for agents like Claude Code and Cursor — four ways into the same catalog of OCR, embedding, speech, and vision workloads. Same interfaces you already use, at a fraction of the cost.

Read the docs →See pricing

pip install commoncompute·npm i @commoncompute/sdk

quickstart.pypython

1# Embed 10M documents with one call
2from commoncompute import Client
3
4cc = Client(api_key="cc_live_…")
5
6job = cc.embeddings.create(
7    model="bge-base",
8    input=open("docs.jsonl"),
9    priority="batch",  # half the realtime price
10    max_spend_usd=50,  # hard cap
11)
12
13print(job.price_usd, job.eta_seconds)  # locked before it runs
14vectors = cc.result(job)

Examples

Idiomatic SDK, one-liners per workload.

Embeddings

Bulk vector generation for search, RAG, clustering.

embeddings.pypython

1# Nightly re-embed of 10M product descriptions
2job = cc.embeddings.batch(
3    model="bge-base",
4    inputs=stream_from_pg("products"),
5    priority="batch",
6)

Migrations

Already running elsewhere? Swap the endpoint.

The native SDK is the recommended path for new projects — but if you have an existing pipeline, two migration routes work today: a boto3-compatible AWS Batch endpoint (full guide) and OpenAI-compatible endpoints for embeddings, chat, and transcription — swap two env vars.

Coming from
What you have
What changes
Coverage

AWS Batch

boto3.client('batch')

same client, endpoint_url swap

submit_job, describe_jobs, list_jobs

OpenAI

OPENAI_BASE_URL + OPENAI_API_KEY

point both at Common Compute

embeddings, chat, transcription

Guarantees

Production-grade, not a hobbyist cluster.

Deterministic quotes

Every job is priced before it runs. The number you see at submission is the number you pay.

Gated execution

Jobs run in native runners on provider Macs, reached only through a signed task assignment and a vetted-input gate. Per-job inputs and outputs are scoped to that job; runner processes and model weights are reused between jobs for warm-start latency. API request logs keep route, status, latency, client IP and user agent for 30 days — never request or response bodies.

Signed receipts

Every completed task emits a cryptographic receipt. Audit trail, attributable to the node that ran it.

Exactly-once semantics

Dedup keys per-task. Reconnect, restart, or re-run without duplicate work or double-billing.

Hard spend caps

Set a max_spend_usd. We pause the job before we cross it, never after.

No lock-in

Compatible interfaces mean you can move back to hyperscalers any time by flipping a flag.

Walkthrough

Watch the first task land.

Walkthrough · 90 sec

Submit your first job in 60 seconds

From npm install to a signed receipt — keys, quote, dispatch, result. Recorded against the live network.

Sign up at commoncompute.ai/signup and add a card.
Copy your API key from the dashboard.
npm install @commoncompute/sdk.
Submit a small job — embeddings is cheapest.
Inspect the signed receipt the network returns.

Run one job. See it land.

No cluster setup, no minimums. A locked quote before every job, billed only on success — you'll know in five minutes whether this is cheaper.

Get started →Read the docs