AI Studio · Desktop app for Claude · macOS, Windows, Linux

Run AI in your real work.
Keep control of every move.

GainWix AI Studio is a local-first control plane for Claude Desktop, Cowork, and Claude Code. One gateway for every tool call, an approval inbox that holds back only the risky moves, and a tamper-evident record of everything your AI did — all on your machine.

Download for macOS See how it works ↓
Also for Windows and Linux
Work with AI that works with you.
Main console — the gateway / overview view
The old way

Your AI can read your files, spend your budget, and act on your behalf — and today you can't see or stop any of it.

GainWix AI Studio is the single place that puts all of it under your control.

Gateway

One entrance. Every call accountable.

Connect Claude Desktop, Cowork, or Code once. Every tool call then flows through a single policy pipeline — authenticated, authorised, rate-limited, audited.

  • One-command connect. No JSON editing, ever.
  • Per-host and per-tool allow-lists — like "database: read-only".
  • Budgets with a soft warning at 80% and a hard stop at 100%.
  • Destructive actions — write, send, delete — denied by default until approved.
Gateway view — connected clients, servers, policy status. Real numbers only.
Workmates studio or the HR Workmate
Workmates

Role-aware teammates, not one-off chats.

Persistent Workmates that work inside your AI apps or run on a schedule with your own key. Anything that acts lands in the Approval Inbox first.

The HR Workmate is the first one available today — a training platform. Author courses with modules, lessons, and quizzes; take and build them from inside Claude; issue completion certificates; run reviewer workflows.

Explore Workmates →
Approval Inbox with a real pending action
Approvals

The risky moves wait for you.

Routine, allow-listed work finishes on its own. Only gated actions and Workmate drafts land in the inbox — approve, edit, or decline, with the full context in front of you.

  • Enforced by the daemon, not the UI — approvals hold even when the app is closed.
  • Decide from a notification, the palette, or the CLI.
  • The product's spine: AI that asks first.
Audit

Know exactly what your AI did at 14:02.

An append-only, hash-chained ledger records every gateway call and every Workmate run. Tamper-evident by construction — and replayable.

  • Replayable traces — the agent debugger your AI apps don't ship.
  • Export the audit with an integrity proof. Pass a security review in an afternoon.
  • Tool outputs tagged and prompt-injection flagged in the ledger.
Audit ledger / a replayed trace
Cost

See spend before the invoice does.

Per-model, per-server cost, metered exactly for API-key traffic and honestly labelled everywhere else. Budgets that actually stop a runaway autonomous run.

  • Exact metering for API-key traffic; honest estimates everywhere else.
  • Consolidated team spend from exported ledgers.
Cost dashboard — real spend figures only, or none
On your machine

Your data never leaves your computer.

Secrets, policies, and audit data stay local. The cloud handles licensing and updates — nothing else.

Licensing

Offline-verifiable licence keys — paid features keep working air-gapped.

Telemetry

Off by default, content-free, public schema.

How local-first works →
Free developer core

The whole governed core, free and headless.

The daemon and CLI are free forever — gateway, policies, budgets, audit, Workmates — with no GUI and no account.

  • Scriptable JSON output and non-zero exits for CI gates.
  • Encrypted backup and restore.
  • Runs headless in Docker.
Clean terminal capture — real commands, real output

One governed home for the AI you already run.

Download for macOS
Also for Windows and Linux
Work with AI that works with you.