Desktop app

The desktop app. Bring the AI you already pay for.

CodeMonster connects your agent to a governed platform over MCP. CodeMonster is invite only while we work directly with early adopters. Request an invite and we will meet, scope what you want to build, and get you productive. No tokens. No credits. No meter. You bring the AI subscription you already pay for.

Bringing this to IT? What the app sends — and does not is written for them.

What you need

Two things: a license login, and the AI you already pay for.

1

An invite and an organization license

One license covers every builder in your organization. You sign in with the account it gives you, and the platform authorizes every operation as you. How the license works.

2

The AI subscription you already pay for

Your agent runs on your own subscription — your model, your keys, your bill. Any subscription that can drive an MCP-capable agent will do; Claude Code is the path we test and document first. How bring-your-own AI works.

Building requires a connection to your AI provider and to the platform. Offline mode is a property of the field apps you build, not of the desktop app.

The first ten minutes

Install. Sign in. Connect your agent. Build.

  1. Install

    Request an invite. We meet, scope what you are building, and hand over the build for your operating system with the setup done together. Then you are at the sign-in screen.

  2. Sign in with your organization license

    Use the account your organization license provides. The app attaches your session token to every request; nothing reaches the platform without it. See how access is controlled.

  3. Connect your agent

    Connect your agent to the platform over MCP. The Claude Code guide has the exact configuration block; other clients follow the same pattern on the MCP hub. Ask the agent to run whoami — it should answer with your identity. Then run list-tenants and use-tenant, or the first write fails with "Tenant context required".

  4. Feed it a prompt

    Start with a real process. The prompt below assembles a working inspections app from governed platform primitives — objects, forms, a grid, a workflow — through named operations. Copy it, change the vocabulary, run it.

    Create a solution called Field Inspections with prefix fsi_. Groups first: Inspectors, Supervisors. Objects: fsi_site, fsi_inspection with photo and signature fields. A form for inspectors, a grid for supervisors, and a workflow that emails the supervisor and generates a PDF when an inspection is submitted. Preview the diff before committing.

    Edits are staged: preview-solution-edit shows an entity-level diff before commit-solution-edit writes once. The control surface lists every operation.

For the person who has to approve the install

What the app sends — and does not.

Written for your IT lead. The full version is on the security hub; the approver's summary is the IT brief.

The app sends
  • Named operations to the platform, each carrying your token — create-object, add-field-to-form, commit-solution-edit. The platform authorizes every call as you.
  • A record of tool invocations, which we keep for audit and quality review.
  • Nothing to your AI provider. Your agent talks to Claude or ChatGPT under your own account; that traffic does not route through us.
The app does not
  • Carry your model conversation to us. Prompts and replies stay between you and your AI provider; we receive only the named operations the agent calls against the platform — and we record those, as stated above.
  • Need inbound network rules. Requests are outbound and authenticated; nothing on your network listens for inbound connections.
  • Hold admin rights on your backend. The agent acts with your permissions, never its own, through named, guarded operations only.
  • Meter, mark up, or resell your tokens. We never touch them.

Every request carries a bearer token validated on every call; an unauthenticated request receives a bare 401. Apps run on managed cloud infrastructure we operate, built primarily on AWS.

No license yet?

We build your use case live, then scope a pilot around it. Your IT lead can ask the hosting, permissions and audit questions while the app is on screen.

Book a demo

Want the arithmetic first?

One organization license. Unlimited builders, unlimited apps, environments included, nothing metered — not even your live apps.

Pricing and TCO calculator

System requirements

What the app expects from your machine.

Exact minimums are being finalized with the desktop builds and will be published here with the installers. What we can say today:

  • A current 64-bit desktop operating system. Supported versions will be named on this page the day the installers ship.
  • An MCP-capable agent on the same machine — Claude Code today — with an active subscription to its AI provider.
  • Outbound HTTPS to your AI provider and to the platform. No inbound rules.

Something your IT team needs answered that is not here? Ask security and we reply in writing.

Questions

Before you install.

Which operating systems does the desktop app support?

We confirm that when we set up your invite, so you get the build you will actually run rather than a list that ages badly. Ask on the invite call.

Does IT need to approve anything to install it?

The desktop app connects your existing AI agent to the platform over an authenticated MCP surface; it does not need inbound network rules or admin rights on your backend. Your IT lead can read exactly what it sends and does not send in the section above and on the security hub.

Is there a trial or a pilot?

Yes. Book a demo first: we build one of your processes live, and a pilot is scoped around whatever comes out of that session, on working software rather than a sandbox.

Every question, one page

Ready when you are

Start with the conversation.

Tell us what you want to build, or which existing solution you want to migrate and modernize. We scope it with you, set you up, and stay with you until the first solution is shipped.

Request an invite See how it works