About Boundless

Open-weight models over an OpenAI-compatible API, billed on prepaid credits with exact per-request costs.

You call open-weight models through an OpenAI-compatible API, and you pay with prepaid credits. Every response includes exact token counts, so you can recompute any charge from the rate card.

How it works#

Writing an agent? You can try without an account.

  1. Sign up. Your account is designed to be active right away. Add a card on Billing to get $1 of free credits to try it.
  2. Create an API key on API Keys.
  3. Point your OpenAI SDK at our base URL. Existing chat-completions code works unchanged.
  4. Spend against your credit balance. Each request returns its token usage and debits the balance by exactly that much. At zero, requests stop.

Zero data retention#

We never store your prompts or outputs, never write request or response bodies to a log or database, and never train on them. We keep what billing needs: token counts, model, timestamp, and cost. The usage page shows those, and they are enough to recompute any charge yourself.

Section 3 of the terms states this in full, along with limits on what we can promise about third-party infrastructure.

Where to go next#

  • About Boundless: who operates this service, what is published about it, and where to ask for more.
  • Quickstart: get a key and stream a first token.
  • Coding Agents & Harnesses: set up Claude Code, Codex CLI, OpenCode and others with one command, or point them at the gateway by hand.
  • Sandbox: run a coding agent in a Vercel Sandbox that is already pointed at Boundless.
  • API Compatibility: the endpoints we serve, and the ones we don't.
  • System One: call Laya for fast scored decisions, and use it to gate a harness's tool calls or route an agent's requests.
  • Credits & Billing: buy credits, set auto top-ups and key budgets, and what happens at zero.
  • Pricing: models and the live rate card.
  • Rate Limits: your team's throughput cap, and how to raise it.
  • Teams: invite colleagues, share credits, and manage roles.
  • FAQ: when to use this API, when not to, and what is not served.
  • Support: how to reach a human.

Machine-readable resources#

  • Setup guide: local setup written for a coding agent to follow: browser sign-in, a new key, and harness configuration.
  • OpenAPI 3.1 spec: typed endpoints, auth, permissions, errors, and response headers.
  • Authentication guide: the API-key flow and what an agent can (and cannot) automate.
  • MCP manifest: the product MCP server and its Streamable HTTP endpoint.
  • llms.txt: a service summary, fit guidance, and links to every docs page.