Developers

An agent belongs in source control.

Name, purpose, runtime, tools, permissions, triggers, approvals, memory scopes, network reach and resource limits — in one declarative file. Reviewable in a pull request, deployable from a pipeline, reproducible.

Read the docs Quickstart

Permissions are reviewed where code is reviewed

A pull request that widens an agent's access shows that in the diff. This is the single biggest advantage of a manifest over configuration in a console.

The manifest is the truth

What executes is the manifest, not a prompt. An agent's behaviour does not drift as a model changes underneath it, and two deployments of the same manifest have the same permissions.

Studio and manifest are the same object

An agent built in the studio has a manifest you can export, and one written by hand appears in the studio. Not two competing systems where one is authoritative.

Refusals are absent, not configurable

There is no manifest key that grants a refused capability. Writing one produces a validation error naming it, at deploy time rather than at runtime.

Roughly what it looks like

Illustrative and deliberately short. The reference is at docs.agentlabora.com, and pretending a marketing page is documentation wastes an afternoon.

An agent, complete

yaml

name: accounts-receivable
purpose: Follow up overdue customer balances
owner: role://finance-manager
trigger: { schedule: 'weekdays 08:00' }
tools:
  - { call: invoice.get, grant: allow }
  - { call: mail.send, grant: allow }
  - { call: invoice.credit, grant: deny }
approvals:
  - when: overdue_days > 30
    decider: role://finance-manager
limits: { emails_per_day: 100, model_cost_month: 120 }
network: tools-only

Where it stops

A manifest declares intent and the platform enforces it, but it cannot describe a capability that does not exist in the registry or grant one that is refused. It also cannot express an approval that another agent decides — approvals name a person or a role, always.

Developer platform

The rest of it.

QuickstartA deployed agent in under fifteen minutes.SDKsPython and TypeScript, maintained in step.APIAgents, runs, tools and events over HTTP.CLICreate, test, deploy, tail logs, roll back.Custom toolsExpose your own service as a permissioned capability.SandboxSynthetic data, mock tools, no production reach.EvaluationsTest suites that run in your pipeline.PublishingShip an agent to the marketplace.DocumentationThe reference, the guides and the API explorer.

Give AI a job.

Build an agent, choose what it can access, define what it may do, and deploy it to work safely across your business.

Build an agent Build with the API

No code required · Full SDKs and API · Stop any agent instantly · [email protected]