Free
Every repo, open source.
$0forever
Install free- Every commit checked first
- The Mind on your machine
- Agents see past failures
- Rules for any chat
Trusted agents. Proven work.
Connect every agent from any chat or MCP client. Each one works to your rules and every change is proven before it lands.
Click to preview
- ConnectPlug in one agent or a whole team. Chats, CLIs and MCP clients all work to the same rules.
- ProtectSet what each agent may touch, run, deploy or spend. Each change lands only once its tests pass.
- ProveEvery result comes with a receipt tied to the exact commit, so anyone can check it later.
Built in for every agent.
Connect once. Every agent gets context, tools, checks, protection and a receipt.
- Context built in. Each task gets only the rules, files and past failures it needs.
- Tools on hand. Checks, file history and memory as MCP tools any agent can call.
- Checks that fit. The tests covering each change run first, in 40 languages.
- All or nothing. A change lands whole once its checks pass, or not at all.
- Protection. Limits on paths, commands, deploys and spend for every agent.
- Receipts. Each result tied to the exact commit. Skipped is never passed.
Checks for this change
- FormattingPassed
- Type checksPassed
- Unit testsPassed
- Integration testsFailed
- Live runNot run
Held: integration tests failed.
Connect one agent or all of them.
One set of rules for every chat, CLI and MCP client, wherever the work runs.
- CLIs and MCP
- Claude Code, Codex, Cursor, Copilot or any MCP client asks which checks a file owes before it acts.
- Chat
- ChatGPT, Claude, Gemini or any chat follows the same rules from one link.
- Tasks
- Each task sets what it may touch, run, spend and deploy, with a time and token budget.
- Repos
- One step adds the commit hook and CI check. Every commit waits for its checks. Free and open source.
- Mind API
- Rates the risk of each step, plans its checks, picks the agent and a backup, and sends critical work to a person. On failure it retries, rolls back or escalates. Free locally; hosted in development.
The checks are free. The Mind API is $10.
Bring your own chat or agents. Checks are free. The hosted Mind API answers before each step: the risk, the checks and who should do it.
Mind API
Hosted, from any computer.
$10/ month
Join waitlist- Everything in Free, plus
- Hosted, reached with one key
- Works from any computer
- Agent handoffs on record
- A receipt for every decision
Pro
For your agents in HeadronOS.
$19/ month
Join waitlist- Everything in Mind API, plus
- All agents in one workspace
- Approve any tool call
- Local, container or remote
- Every action logged
Team
For a team and its agents.
$39/ seat / month
Join waitlist- Everything in Pro, plus
- Rules synced to every seat
- Approval before deploys
- All agent work in one view
- Audit export for security
Enterprise. The Mind API inside your network, with your hosts, models and policies.
Talk to usFlat prices, no metering. Agent subscriptions and model tokens stay with your own provider. Paid plans are in development and not available to purchase yet.
Get early access.
The Mind API opens first. We write once, when your plan is ready.
Questions.
What is Headron?
The control layer for AI agents. You set what each agent may do. Headron checks every change it makes, holds what fails and records what passed.
What does the Mind API do?
It decides before your agents act. It rates the risk of each step, picks the agent for it with a backup, plans the checks it owes and sends critical work to a person. When a step fails it finds the cause and retries, rolls back or escalates. The hosted plan works from any machine, CI runner or cloud agent and keeps one memory for all of them.
How is this different from running my tests?
Tests say whether they passed. Headron asks whether anything tested what just changed and holds the work until something has.
Can I set my own rules?
Yes. Each task can name the paths an agent may touch, the commands it may run, whether it may deploy, spend money or go online, and a budget for time and changes. Work outside them is stopped.
Which agents and chats does it work with?
Claude Code, Codex, Cursor, OpenCode, Hermes, GitHub Copilot, Google Antigravity and any MCP client. In ChatGPT, Claude, Gemini, Grok or any other chat, one link gives the assistant the same rules.
What is HeadronOS?
The workspace in the Pro plan, in development. Connect one agent or several, such as Claude Code, Codex, Cursor, OpenCode or Hermes, with local or hosted models. Run work on your machine, in a container or on a remote host, approve each tool call and keep every change, check and receipt in one place.
Can I buy a plan today?
Not yet. The free core works today. Join the waitlist and we will write when your plan opens.
Try it on your project.
Free and open source under MIT. Runs on your machine and in CI. The README has the three install steps.