gusto

One place for every AI agent your team needs.

Run Claude Code, Codex, Pi, OpenCode and other agents with the same controls for credentials, models, spend and oversight.

No sign-up. A 30-minute isolated demo, deleted when it ends.

Use the agents your team already wants.

Switch agents without rebuilding credentials, model access, spend controls or activity tracking. No proprietary agent framework required.

How it works

  1. Pick an agent

    Choose the one for the work.

  2. Give it a workspace

    Files, tools and context stay together.

  3. Keep the same controls

    Keys, models, network and spend stay outside the agent.

  4. See what happened

    Calls, changes, cost and results are recorded.

One workspace. One set of controls. One record.

Tested across multiple agents.

Same task. Same model. Same environment.

AgentFinishedTime
PiYes28 s
Claude CodeYes30 s
CodexYes35 s
HermesYes40 s
OpenCodeYes2 m 46 s

A one-line fix, checked by a test, on DeepSeek Flash. Test-environment measurements from 2 October 2026. They show that each agent runs here, not which is best for your work.

Built for teams that use more than one AI agent, and need credentials, model access, spend and activity to stay under control.

Questions

What gets uploaded?

Nothing from your computer. A demo starts with an empty workspace, and the agent sees only what you type to it.

Which agents and models work?

Claude Code, Codex, OpenCode, Pi and Hermes each finished the test above. The public demo offers all five: OpenCode, Pi and Hermes on open Gemma models, and Claude Code and Codex on DeepSeek Flash. A deployment of your own runs them on the models your organisation connects.

What happens when the demo ends?

The organisation, its workspaces and everything in them are removed. We keep the name, email and company you gave so that we can follow up about the demo, and nothing else.

Who can see my workspace and runs?

A demo organisation is yours alone, on an address of its own. Within an organisation, administrators see people, spend and the record of runs.

Use different AI agents without setting up each one from scratch.