Skip to main content
Try it free Book a call

Halfcycle · For enterprise

AI-built software your organisation can sign off

Your engineers work one way on every seat. Changes are checked as they are written, approvals carry a name, and each project has a page showing its cost. We set it up with your teams and stay until they run it without us.

Book a call Governance and data
Agents run on
Your machines, your accounts
Sign-in
Your identity provider
Checks run
With us, or inside your perimeter under contract
Training on your work
None

What engineering, security, finance and audit each need to know

The head of engineering

Will this make my teams quicker, or only busier?

  • A new colleague needs an invitation and one command. After that they are on the project with the role you chose, and there is nothing else to install or keep up to date.
  • Every change is checked in the session while it is being written, and anything stopped comes with the reason. That leaves reviewers free to read for design.
  • On a repository you already have, it reads the code first and writes the plan from what it finds, showing the line behind each statement. Several repositories can make up one project.

The person who owns the risk

Where does our code go?

Your code stays in your repositories. The agents run inside your own Claude Code or OpenAI Codex, and Halfcycle has no access to the code. When the console points at a file it links to your own remote, under your own permissions. Your Claude or OpenAI credential is never held or proxied, so nothing can be spent on your behalf.

Members sign in through your identity provider, and removing someone from your directory removes their access. Some checks run away from the developer's machine, so the session that did the work is not the one judging it. Under contract, those checks run inside your own perimeter.

Halfcycle does not scan your dependencies, model threats or test for intrusion. Those stay with your security team.

The person who pays for it

What did that feature cost us?

Cost
The console shows what each phase has cost in money and how long it has taken, for one project and for all of them together.
AI usage
The agents run on your own Claude or OpenAI account, as part of your normal usage with them.
Seats
You pay for members who write code. A lead, a product manager or a client who only reads holds no seat.
Billing
Enterprise is by contract and billed by invoice.

The person who has to show the working

Can we show who decided what?

Two things show it: the phase record and the audit log. The record is written when a phase closes and covers the plan, what was built, each decision with the options considered, and who approved what. It cannot be changed afterwards, and a copy sits in your own git history. The audit log is a dated list of invitations, rule changes and phase approvals.

Agents cannot give those approvals. Each project names the people who can, and a second approver can be required.

Halfcycle is not certified against any standard and cannot certify what you build. An auditor gets the phase records and the audit log.

Rules your organisation writes apply to every project

An Admin writes a convention, a check or a required step once. It applies on every seat from the next edit, and when it fires it says which rule it was. An organisation's rule can make the checking stricter. It cannot remove a check or weaken one.

Organisation ruleExample

“Any change under services/billing needs a second approver before its phase can close.”

Written by
D. Osei, Admin · 4 Aug
Applies to
Every project, from the next edit
Last fired
Phase 4, Billing export · held for a second approver

How it rolls out, from a first call to every team

Each step is a fixed fee, quoted on request.

  1. Two weeks

    The AI-Build Audit

    We read your repositories, your review process and your spend. You get a written finding and the order to change things in.

  2. Four weeks

    The adoption pilot

    One team ships real work on your repositories for four weeks. We lead at the start and your engineers lead by the end, and you leave with measured cost and time from your own work to show your leadership.

  3. Scoped with you

    Implementation

    Your rules, single sign-on, your tracker and training, across more teams than a pilot covers. We scope it with you after the pilot.

What we build on an engagement belongs to you. Our people use the roles you give them and nothing longer-lived, and their access ends with the engagement. We do not bill by the hour. When an engagement ends, your organisation is on the Team or Enterprise plan.

Questions procurement asks

Do you need access to our repositories?

No. The product never has it. On an engagement our people work inside your organisation under the roles you give them, and that access ends when the engagement does.

Which AI coding agents does it work with?

Claude Code and OpenAI Codex, on macOS, Linux and Windows. Your engineers keep the tool and the account they already have.

What do we keep if we stop?

Your code, your specifications and every phase's record are plain files in your own git history, and they stay yours. What stops is the service: new sessions no longer get the plan, the rules or the checks.

Can we try it before talking to anyone?

Yes. Start a team trial from the console, invite your engineers by link and run it on a real repository. It needs no card.

Book one call for all four

Bring your security and finance leads. We work through one of your repositories and they ask their own questions.

Book a call Start a team trial