Build with the API
What you can do with the AgentWorks API and how the pieces fit together.
The AgentWorks API lets your own systems work with one workspace: start agent runs, create and follow tasks, search knowledge, send chat completions and read usage analytics. Everything the API does is done as a person in that workspace, with a key that you scope to exactly what the integration needs.
What you can build
Hand work to agents
Create a task from your ERP or CRM, assign it to an agent and follow the run.
Webhook triggers
Start an agent or workflow from an event in another system, with a signed request.
Chat completions
An OpenAI-compatible endpoint: change the base URL and the key, keep your code.
Ask an agent
Send a message to one of your agents and get its answer back.
Search knowledge
Retrieve the most relevant passages from your knowledge bases.
Analytics
Usage and spend per agent, user and team, plus balance and plan usage.
Agencies
Read the monthly usage of the clients you manage.
The basics
| Base URL | https://api.agent-works.ai |
| Authentication | Authorization: Bearer aw_ak_… (details) |
| Format | JSON over HTTPS |
| Permissions | Scopes per key |
| Limits | Rate limits and errors |
Every endpoint a key can reach is listed in the API reference tab, each with the scope it needs.
How access works
A key belongs to one workspace and acts as the person who created it. It can never do more than that person's role allows, and it stops working when that person is deactivated. On top of that, the key only reaches the endpoints its scopes allow.
Two things are never possible with a key, whatever its scopes:
- Deciding approvals. When an agent or workflow asks a person to approve an action, that decision stays with a person in the app.
- Changing guardrails. Approval settings, per-action confirmations and the personal-data policy are changed in the app, not over the API.