Standard Compute is an OpenAI-compatible API with unlimited frontier-model compute at one flat price — from $9/mo, no per-token billing, no 429s. It works as a custom model provider in any agent or tool — Codex CLI, OpenCode, Cursor, Claude Code, Aider and more. Point the base URL at us, drop in your key, set the model to standardcompute, and run — no code changes, no per-token billing.
Get your API key →Three settings connect Standard Compute to anything OpenAI-compatible.
Exact setup — the custom OpenAI-compatible base URL and where it goes — for the tools developers connect most.
Pick your agent for its exact setup. Standard Compute powers coding agents, personal automation agents, and conversational assistants alike.
Anthropic's terminal coding agent — plans, edits, runs tests, and ships multi-file changes with frontier-model quality.
Nous Research's open-source self-improving agent — persistent memory, a learning loop that creates reusable skills, and access from 20+ messaging platforms.
OpenAI's open-source terminal coding agent — sandboxed local execution with approval modes, powered by GPT-5-class models.
Google's open-source terminal AI agent with a 1M-token context window and the most generous free tier of any frontier agent.
The viral open-source autonomous agent that runs on your own machine and acts through your messaging apps — 100+ skills spanning browser, email, files, and APIs.
VS Code extension that provides AI-powered coding assistance with multi-file editing and refactoring.
AI-first code editor built on VS Code with deep codebase understanding and natural language editing.
GitHub's AI coding assistant with inline suggestions, chat, and deep GitHub ecosystem integration.
The open-source agent that defined the VS Code agent category — Plan/Act approval modes, MCP support, and full bring-your-own-key freedom.
AI-powered code editor (formerly Codeium) with Cascade flow for multi-file, multi-step coding tasks.
Terminal-based AI pair programmer that edits code in your local git repo with any LLM.
Open-source AI coding assistant for VS Code and JetBrains with customizable models and context.
Terminal-based AI coding agent with a clean TUI, multi-provider support, and LSP integration.
Mario Zechner's radically minimal open-source terminal coding agent — a handful of core tools, a fast TUI, and TypeScript extensions for everything else.
A coding-first fork of Pi with an IDE wired into the terminal — hash-anchored edits, LSP navigation, real debugger control, and subagents on a Rust core.
The power-user fork of Cline — a VS Code agent with switchable modes (Architect, Code, Debug, custom), deep configurability, and BYO API key.
Cognition's autonomous software engineer — takes a ticket, plans, codes, tests, and opens a PR in its own cloud environment, managed from Slack or the web.
Sourcegraph's agentic coding tool — thread-based, aggressively autonomous, always running frontier models at full reasoning in your editor or CLI.
Google's asynchronous coding agent — connect a GitHub repo, hand it tasks, and get tested pull requests back from cloud VMs while you do something else.
Google's agentic IDE built around Gemini 3 — an agent manager that plans, codes, and verifies its own work in the editor, terminal, and a controlled browser.
Building your own agent, or using an SDK? If it speaks the OpenAI format, it works. Use these three settings anywhere.
The connection is the easy part. What you get on the other side is unlimited frontier-model compute at one flat price — built for agents that run all day.
Almost certainly. Standard Compute exposes an OpenAI-compatible API, so any agent or tool with a custom OpenAI-compatible provider setting works — Cursor, Cline, Aider, OpenCode, Codex CLI, Continue, Kilo Code, OpenClaw, Hermes Agent, and more. You only change the base URL and API key. The exceptions are tools locked to a single vendor's API with no custom-provider setting, like Claude Code.
Not directly. Claude Code is built for Anthropic's Messages API and has no native OpenAI-compatible base-URL setting. Community router projects can translate between the two formats, or you can get the same flat-rate setup natively in an OpenAI-compatible terminal agent like OpenCode, Codex CLI, or Aider. Our Claude Code compatibility guide covers what actually works.
Set the base URL to https://api.stdcmpt.com/v1, paste your Standard Compute API key, and set the model to "standardcompute". Our router picks the best underlying model for each request. No code changes beyond those three settings.
No. The request and response format is identical to OpenAI's. If your agent or SDK already speaks the OpenAI format, you only swap the base URL and key.
Per-token billing makes always-on agents unpredictable and expensive. Standard Compute gives you unlimited compute at one flat monthly price, automatically routes each request to the best frontier model, and keeps the exact OpenAI request and response format — so you swap the base URL, not your code.
Always "standardcompute". It is a single model identifier that routes to the best available frontier model for each request, so you never have to pick or update model names.