## The Respan Agent

The Respan Agent is the AI workspace inside the Respan platform. Describe what you want to investigate or configure, add the relevant context, and continue the conversation as you refine the result.

## What you can do

| Goal | Example request |
| --- | --- |
| Investigate telemetry | `Explain why trace [trace_id] is slow.` |
| Analyze a group of logs | `Run a root cause analysis for suspicious logs from the last 7 days.` |
| Work with prompts | `Create a customer support prompt.` |
| Configure monitoring | `Create a monitor when cost exceeds $100/hour.` |
| Prepare an evaluation | `Build a hallucination evaluator.` |
| Search telemetry by content | `Find logs relevant to "rate limit exceeded" in the last 24 hours.` |

The Agent can help you move between these workflows, but you should review the result and confirm the referenced project, time range, filters, and settings before relying on it.

## Full-text search

Ask the Agent to find spans by what they contain, not only by structured fields like model, status, or cost. Full-text search matches the text of each log, so you can locate a request from a phrase a user typed or a string your app returned.

- `Find logs relevant to "payment failed" in the last 7 days.`
- `Search for traces that mention "checkout" and returned an error.`
- `Show me spans that contain order id 8842.`

The Agent runs this as a **relevant to** filter over your logs’ full text, backed by a text index so it stays fast across large volumes. You can apply the same **relevant to** filter yourself when you [explore logs](/content/docs/documentation/features/observe/explore-logs/index.html).

## Start a conversation

### Open the Agent

Select **Chat** in Respan or open the [Agent page](https://platform.respan.ai/platform/agent).

### Describe the outcome

Enter your own request or select one of the starter workflows. Include useful identifiers, such as a trace ID, prompt, time range, or metric threshold.

### Add context

Select the composer **+** ( **Add context or attachments**) when the Agent needs a file, a saved skill, or a resource from the platform.

### Review and continue

Inspect the response, follow any links back to the underlying Respan data, and ask a follow-up question. Select **New chat** in the sidebar or the **+** in the header when you want to start with a separate conversation context.

## Add context

The plus button in the composer opens three context paths:

| Context option | Use it for |
| --- | --- |
| **Add images or documents** | Attach a file that is relevant to the current conversation. |
| **Saved skills** | Apply reusable instructions that your organization has saved for the Agent. |
| **Reference context** | Select relevant resources from Respan instead of describing them only in text. |

Add only the context needed for the task, and verify that you selected the intended project resource before sending the request.

## Create a reusable Agent skill

A conversation contains the instructions and follow-ups for one chat. An Agent skill saves reusable instructions so people in your organization can apply the same workflow from the composer.

### Open Agent skills

Go to [Settings → Agent skills](https://platform.respan.ai/platform/api/agent-skills).

### Create the skill

Select **New skill**, then add a clear name, a short description, and the instructions the Agent should follow.

### Add reusable context

If the workflow depends on a prompt, grader, or monitor, use the plus button or type `@` to attach it to the skill.

### Save and use the skill

Select **Create skill**. In an Agent conversation, open **Saved skills** from the context menu and choose the skill you want to apply.

Write instructions around a repeatable outcome. For example, a trace-triage skill can ask the Agent to identify the slowest spans, summarize likely causes, and link to the relevant logs.