Agent
An Agent configuration defines how to connect to the service under test and what inputs it requires. An evaluation task calls the service for each Case according to the field mapping, then uses evaluators to assess the results.
For service maintainers and evaluation users. Verification baseline: the project's current implementation as of 2026-10-30. Entry points in deployed environments may vary by version and permissions.
Search keywords: AE Agent, A2A, HTTP, OpenClaw, Hermes, Connector, input variables.
1. Connect an AE Agent
- Go to Implementation → Agent, click Create Agent, and select AE Agent (A2A 1.0).
- Enter the Agent Name and check the automatically populated Service URL. The current Agent form allows you to change the URL for your deployment. Reload the catalog after changing it.
- Select an AE Agent and an available model from the catalog. You cannot complete the configuration until the catalog is loaded and a model is selected.
- With TE SSO, your current login credentials are used automatically. In other authentication environments, select credentials as required by the page.
- After saving, test connectivity, then verify actual calls with a few Cases. Each evaluation Case creates a separate Context through A2A.
2. Connect an HTTP Agent
- Select HTTP as the connection type, then enter a name, service URL, and required input variables. Separate multiple variables with commas as instructed in the UI.
- Select the credentials required by the service and save. The service maintainer should confirm that the API meets the platform's HTTP JSON integration contract. An arbitrary HTTP URL alone does not mean the service can be evaluated directly.
- After testing connectivity, bind all required inputs in the field mapping when creating a task.
3. Connect OpenClaw or Hermes
| Type | Service prerequisites | Common configuration |
|---|---|---|
| OpenClaw Gateway | Enable the Chat Completions HTTP API. | Confirm the Agent ID and Gateway authentication. The default is the main Agent; configure other Agents or password authentication in Advanced Settings. |
| Hermes Agent | Enable the Hermes API Server. | Confirm the Agent ID, Profile, and connection method. Use Bearer credentials for API_SERVER_KEY. |
For enterprise private networks, you can use a customer-side Connector in the corresponding Advanced Settings. Hermes offers both direct platform connections and customer-side Connectors. The environment maintainer should provide the Connector URL, TLS mode, and certificate file names. Use a configured secure connection in production. A certificate file name shown in the UI does not mean you can paste the private key content there.
4. Edit, test, and understand limitations
- Open Edit from the Agent list and check the URL, type, inputs, and authentication. Save your changes and test again.
- If saving fails, correct the reported error before retrying. Once connectivity passes, validate with a small set of samples in the actual format.
- Run history uses frozen configurations. Editing an Agent does not automatically rerun past evaluations or produce new results.
The current creation options are AE Agent, HTTP, OpenClaw, and Hermes. Although a compatibility implementation exists for the OpenCode type, this version does not cover creating OpenCode Agents. The detail available in execution traces depends on the integration protocol and source evidence. A successful connection does not guarantee a complete Trace.
5. Next steps and related pages
- Bind inputs and run: Create evaluation tasks and map fields.
- Prepare authentication for external services: Credentials.
- Understand execution evidence: View results and execution traces.
Return to the manual home page:

