Standard Compute documentation
Standard Compute gives your AI agents one API key for frontier and open models. You pay a fixed monthly plan, from $19 a month, instead of per-token bills, and a router picks a capable model for each request. Each plan has a monthly compute budget: requests run at full speed until it is used, then stop with 402. There are no overage charges and no 5-hour or weekly windows.
The API is compatible with both OpenAI and Anthropic clients. In most tools, you change the base URL and API key and everything else keeps working.
Quickstart
Make your first request in under a minute.
Connect an agent
Claude Code, Codex, OpenCode, OpenClaw, Hermes and more.
Models and routing
Use smart routing, or pin an exact model.
API reference
Endpoints, parameters and errors.
At a glance
| Setting | Value |
|---|---|
| Base URL (OpenAI clients) | https://api.stdcmpt.com/v1 |
| Base URL (Claude Code / Anthropic clients) | https://api.stdcmpt.com |
| Authentication | Authorization: Bearer YOUR_API_KEY |
| Smart routing model | standardcompute |
| Endpoints | Chat Completions, Responses, Messages, Completions, Models |
| Context window | Up to 1M tokens, depending on the model |
| Plans | From $19/month, each with a fixed monthly compute budget. See Budgets and limits |
How it works
- Your agent sends a normal OpenAI or Anthropic request to Standard Compute.
- If the model is
standardcompute, the router picks a model from the ones you've enabled on the Models page. If you send a specific model ID, that exact model is used. - The response comes back in the format your client expects. Usage is counted against your plan's monthly budget, which you can watch in the dashboard.