.png)
What you can do with the Agent Platform API
You can use our endpoints to perform the following actions:- Create trajectories: Start a new run by defining an agent and an objective.
- Monitor progress: Retrieve a trajectory’s current status, events, and changes in real time.
- Interact with agents: Send messages, commands, or simulated user actions.
- Access resources: Fetch data and artifacts generated during a run.
- Manage agents: List available agents and their identifiers.
- Submit feedback: Report bugs, issues, or suggestions directly from your integration.
Key concepts
This section covers the main concepts and objects that shape how the Agent Platform API works:| Concept | Description |
|---|---|
| Trajectory | The lifecycle of an agent pursuing an objective. Trajectories move through statuses such as pending, running, completed, or failed. |
| Agent | A worker capable of executing a trajectory. |
| Objective | The goal or task you assign to an agent, for example, “find the most highly rated hotels in Paris” or “create a list of meeting objectives.” |
| Interaction | User inputs that can influence a running trajectory, including messages, commands, and browser-like actions. |
| Events | Logs of what happened during a trajectory, including agent actions and state changes. |
Use cases
This section provides a few of the many potential use cases the Agent Platform API can fulfill:| Use Case | Description |
|---|---|
| Workflow automation | Automate web actions, like navigating pages, filling forms, or clicking buttons. |
| Data collection | Gather structured information from web pages or resources in a trajectory. |
| Process monitoring | Track the progress and status of automated tasks in real time. |
| Interactive simulations | Simulate user interactions with web applications for testing or experimentation. |
| Event-driven triggers | View trajectory events, such as errors, completions, or specific agent actions. |
Get started
Follow these steps to start using the Agent Platform API:1
Generate an API Key
Log into PortalH or see our Authentication guide to create your API key.
2
Create a trajectory
Define a task for an agent by specifying an objective and any relevant parameters such as agent type, visibility, or timeout. Submit this to start the trajectory.
3
Monitor the trajectory
Track the status of your trajectory, check for updates or view events such as
ChatMessageEvent or AgentStartedEvent to see what your agent does during a run.4
Modify the trajectory (Optional)
Send messages, commands, or execute actions to influence a running trajectory.
5
Retrieve outputs or resources
Once the trajectory completes, fetch any generated resources or results. This could include logs, files, or other outputs captured during the run.