Test end-to-end

Run the tool-using agent end-to-end, verify autonomous API calls, and confirm the full agent pipeline.

With the tool connected and the agent contract updated, you are ready to test the full pipeline.

Step 5 - Test end-to-end

To test the agent end-to-end, use the Debug button from the toolbar at the top of the agent builder. This opens up the Debug configuration dialog, then do the following:

1. Open the Entrypoint arguments tab
2. Enter a quest description from the code block below
3. Then select Save & Debug to run it

The villagers report a massive creature has been destroying farms on the edge of the forest at night.

After you click on Save & Debug, you can watch the Execution Trail at the bottom of the agent builder as it executes. The Execution Trail shows the agent's full decision process. For a tool-using agent, expect at least three events:

1. The agent's initial analysis of the quest
2. The tool call with the search term it chose
3. The tool response with the candidate list

The agent then reasons over those candidates before producing its final output.

You should see:

1. The agent calls Monster Query with a search term it chose based on the quest.
2. The tool returns a list of candidates.
3. The agent selects the best match and returns the five output fields.

Select the HTTP Request - Open5e Monster Query span to see the request the tool actually sent: the v2/creatures/ URL, the search term the agent chose in name__icontains, and the document__key and fields values you configured in Step 1. The response below it carries the key identifier for each candidate.

Verify the output contains all five fields: monsterIndex, monsterName, monsterType, monsterCr, and monsterReasoning. monsterIndex is the API's key, so it carries the srd_ prefix - for example srd_giant-ape. The monsterReasoning field should explain why the agent chose this monster for the quest.

---

What you built

You have built the first component of UiPathfinder:

• Built a live API connector that searches the 5e SRD: one input, one HTTP request, one output.
• Connected it to your agent as a tool with a single selection in the agent builder.
• The agent now decides autonomously when to call the tool, what to search for, and which candidate best fits the quest.

The two-step reasoning chain (search term selection and candidate evaluation) is exactly the kind of trajectory that produces meaningful evaluation signal; this is what the next lab is about.

What's next

• Open5e API docs - explore the full monster search API to understand what fields are available.
• UiPath Integration Service - build tools that connect to SaaS APIs without writing a custom workflow.
• UiPath Community - forums, how-tos, and developer discussion.