Standard Compute

Setup guides · 24 tools

How do I connect my AI agent to Standard Compute?

Set three values in your agent’s custom OpenAI-compatible provider: the base URL https://api.stdcmpt.com/v1, your Standard Compute API key, and the model standardcompute. Claude Code, Codex CLI, Hermes Agent and OpenClaw connect with the Standard Compute launcher instead: run standardcompute launch <tool> and approve the browser sign-in.

Get your API key

The three settings

Base URL

https://api.stdcmpt.com/v1

API Key

Your Standard Compute key

Model

standardcompute

The same values work in the OpenAI Python and Node SDKs (pass base_url / baseURL), LangChain and plain curl. SDK examples →

Where the settings go, tool by tool

Each guide shows the exact menu or config file. Tools marked Launcher use standardcompute launch instead.

Not listed? Any tool with a custom OpenAI-compatible provider setting takes the three settings above.

More than a base-URL swap

One monthly price
Plans from $19 a month, each with a fixed compute budget you can watch live in the dashboard. No per-token bill and no 5-hour or weekly windows. When the budget is used, requests stop until renewal or an upgrade.
Two API formats
OpenAI-compatible Chat Completions and Responses at /v1 for most tools, and the Anthropic Messages API for Claude Code.

Integration FAQ

Does Standard Compute work with my AI agent?

Yes, if the agent has a custom OpenAI-compatible provider setting. That covers Cursor, Cline, Roo Code, Aider, Continue, OpenCode, Kilo Code, Zed, Trae, Pi, Oh My Pi, Amp, goose, Qwen Code, Factory Droid, Junie CLI, GitHub Copilot, Crush, Kimi Code CLI and Warp. Claude Code, Codex CLI, Hermes Agent and OpenClaw connect through the Standard Compute launcher. Tools that only accept their vendor's own models cannot be pointed at Standard Compute.

Does Claude Code work with Standard Compute?

Yes. Standard Compute serves the Anthropic Messages API (/v1/messages). Run standardcompute launch claude, or set ANTHROPIC_BASE_URL to https://api.stdcmpt.com (no /v1) with your Standard Compute key and ANTHROPIC_MODEL and every ANTHROPIC_DEFAULT_*_MODEL slot set to catalog IDs. Requests draw on your plan's monthly compute budget.

Do I need to change my code?

No. If your agent or SDK already speaks the OpenAI format, you change the base URL and the key. Some tools add their own limits on custom providers; each setup guide lists them.

What model should I select?

Use "standardcompute" to let routing pick a frontier or an efficient model for each request. To pin one model, use its exact ID from the Models page.

Why route through Standard Compute instead of paying per token?

Per-token billing makes always-on agents expensive and hard to predict. Standard Compute charges one monthly price with a fixed compute budget, routes routine steps to efficient models, and keeps the OpenAI and Anthropic request formats, so you change the base URL, not your code.