# Delegate coding to your AI agent | Codemind

## Delegate the coding. Pay only when it works.

Your coding agent is good, but every file it writes itself spends your tokens and your attention, including the routine ones. Codemind takes that work off it.

Codemind is a coding service your AI agent calls. You give it a task and acceptance criteria, and it returns code that has been tested against those criteria. If the build doesn't pass, you aren't charged.

## What is Codemind?

Codemind is a build service that Claude Code, Cursor, Codex and OpenCode can call. Your agent sends a title and acceptance criteria. Codemind writes the code and tests it against your criteria before anything comes back.

Say you need a rate limiter on your API. You write three sentences of criteria and your agent sends them. A few minutes later the files come back, your agent writes them into the project, and you get on with the work only you can do.

See the full picture on the [home page](/), or read what changed recently in the [changelog](/changelog/). Not sure which tasks to hand off? Read [what to hand off and what to keep](/guides/what-to-hand-off/).

## How it works

- ### Set it up with one prompt Paste the setup prompt from the home page into your agent. It writes one script for you to run, which creates a free account and connects Codemind. No sign-up form and no card.
- ### Send a task A title, the acceptance criteria and the stack. The more specific the criteria, the better the result.
- ### Get working code, or a clear failure A build that passes comes back as files for your agent to write. A build that fails comes back with an error code and costs nothing.

## What does it cost?

You get 100 verified builds free in your first month, then 50 a month, with no card. Builds beyond that are $0.10 each, charged only when a build is verified. See the [pricing section](/#pricing) for the details.

Paid top-ups are coming soon.

## Why we built this

We kept watching agents burn their context on routine coding. We wanted to take that work off the agent and make the result dependable enough to build on. Codemind does the coding, and you decide what gets built. More on [who we are](/about/).

## Questions

Which AI coding tools work with Codemind?Claude Code, Cursor, Codex and OpenCode.

What do I pay when a build fails?Nothing. You pay only for verified builds, and builds inside your free allowance cost nothing either way.

Do I need an account first?No. Your agent can set up a free account from one prompt. You run one script and it stores your key for you.

Does Codemind replace my coding agent?No. Your agent still plans the work and edits your project. Codemind takes the tasks it delegates and returns code that has already been tested.

## Delegate your first task. See what comes back.

[Start free](https://app.codemindhq.dev)