Solutions · Use cases
Hanzo for AI agents
Define an agent once — model, instructions, tools and a budget — run it by name from anywhere, and read back every step it took.
What it does for you
Define once, run by name
An agent lives in the organization, not in a script. Call it from the API, the CLI, a workflow or chat; each run is recorded with its input, output, model and duration.A budget it cannot pass
Every session carries a cap in micro-dollars. Raise it, lower it or remove it, and read what each agent has spent.Every operation is a tool
Each subsystem’s operations are exposed over MCP, so an agent reaches the same surface a person does — under the grants its credential carries, and no more.A computer when it needs one
The agent leases a sandbox, runs commands, drives a desktop or your signed-in browser, and gives the machine back.Started by events
Fire a run from a webhook, a schedule or a message, and let a durable workflow carry it through restarts.What it runs on
The products underneath, on one account, one key and one bill.
Agents
Define an agent once — model, instructions, tools and a budget — and run it by name.Hanzo Bot
A teammate with its own computer that signs in to your tools and reports back.MCP
A shell, a filesystem, a code index, git and HTTP for any MCP client.Agents SDK
A Python SDK for agents that hand work to each other behind a router.Hanzo Auto
Workflows fired by a webhook or a schedule, with runs that survive a restart.Hanzo Computer
An agent that drives any app by screenshot, mouse and keyboard.Security and compliance
Credentials narrowed to the job
Access is a grant at a path — organization, workspace, project — and a credential handed to an agent can be narrowed to exactly what it needs.An audit trail you can query
Every request is a row: the organization and user who acted, the address they came from, what they called and what the server answered. An agent’s run is recorded the same way.More use cases
Put it to work
Start in the builder today, or talk to us about your organization.