enterprise

AI coding your security team can sign off on.

AdultCoder puts pro-grade coding agents in your engineers' hands without sending your code to someone else's cloud, scattering secrets into prompts, or locking your team's knowledge in personal accounts.

01

Deploy where your policy already lives

On developer machines, on model servers inside your own network, or with hosted models through your own keys, chosen model by model.

02

Give engineers real leverage

Pro-grade agents for the people already trusted with critical systems, on a proprietary native Rust engine that stays light on large codebases, with brakes before anything destructive and a budget on every turn.

03

Keep compliance explainable

Machine policies a project can't loosen, permissions kept where they were granted, and records anyone on the review can read.

who it's for

Built for the teams already carrying production risk.

  • A
    Product teams under compliance rules

    Teams that want AI speed with data kept in place, approvals and an audit trail.

  • B
    Platform and security engineering

    Groups that roll out tools across many teams and answer for how they behave.

  • C
    Long migrations and modernization

    Work that runs for hours, with budgets, stop rules, and an agent that waits out an outage.

what your review gets

  • 1
    A clear data map

    Where code, prompts, keys and logs live, and which provider saw what.

  • 2
    No hidden state

    Permissions are project-local, and machine policies apply to every project.

  • 3
    Evidence on request

    Hash-chained audit files exported with a manifest, and sessions that can be walked through step by step.

where it's going

One engine and one policy, across every tool your teams use.

The terminal agent and the Mac app come first, in a beta soon. The rest of the family is in development on the same engine and the same controls, so a policy you set once follows your teams everywhere.

ac and AdultCoder for Mac

beta soon

The coding agent in your terminal, and a menu bar app for keys, model servers, defaults and policies.

AdultCoder Desktop and IDE

in development

A desktop app for the whole team, and an editor built around the agent.

Windows and a private model server

in development

The ac command and a tray app for Windows, and open-weight models served from your own hardware to your team.

enterprise FAQ

The questions legal, security and engineering leadership ask.

A tool earns its place by surviving scrutiny, not by its demo.

Yes. Open-weight models through Ollama on each machine, or llama.cpp, vLLM or LM Studio servers in your own network. Hosted models are an option you turn on, not a requirement.
Yes: Anthropic, OpenAI, DeepSeek and Qwen, each with your own key in the system Keychain, used only when a model names it. Work never moves to another provider's model without asking.
Yes. ac audit export copies the audit files out with a manifest of their hashes, and ac audit verify checks their chains. Sessions can be walked through step by step with ac replay.
Project memory is a plain file in the repository that every model reads. Sessions export, or sync through a shared folder, signed and encrypted, and resume on any model.
Yes. Persisted grants to another repository live in the asking project's own preferences, not globally and not in the repository it reads.

private briefing

Tell us your constraints. We will show you the controls.

Your environment, your data rules and how your teams work: one conversation tells us both whether there is a fit.