> For the complete documentation index, see [llms.txt](https://docs.coherent.global/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.coherent.global/assistant/agent/chatting-with-agent.md).
# Chatting with Agent
Agent uses a chat interface. You can send a message, attach files, ask follow-up questions, and guide a task step by step.
Agent chat interface
A good Agent conversation usually starts broad enough to explain the goal, then narrows into specific sheets, ranges, checks, or edits. You do not need special commands, but clear instructions produce better results.
## Write effective prompts
Tell Agent what you want, where it should look, and how detailed the answer should be.
Use this pattern:
> Do \[task] using \[sheet, range, workbook, or attachment]. Return \[format or level of detail].
Examples:
* "Analyze `Revenue!A1:H120` and summarize the key trends in bullets."
* "Review the active sheet and identify formulas that look inconsistent."
* "Create a pivot table from the selected range with Region as rows and Revenue as values."
* "Compare the attached workbook against the open workbook and list material differences."
When a task has risk, state the limits directly:
> Review the model for issues, but do not change the workbook.
> Preview the proposed edits before writing anything.
## Ask useful follow-up questions
You can continue the same conversation with follow-up requests. Agent remembers earlier messages in that conversation, so you can refine the answer without restating everything.
Examples:
* "Show the same analysis by quarter."
* "Explain the second issue in more detail."
* "Which formulas drive the final output?"
* "Apply that formatting to the rest of the table after I approve it."
For longer work, ask Agent to proceed in stages. For example, ask for a summary first, then ask for the top risks, then ask for suggested fixes.
## Control the answer format
If you need a specific format, ask for it. Agent can usually adapt its response to a short summary, a checklist, a table, or a step-by-step explanation.
Examples:
* "Return a table with columns for Sheet, Cell, Issue, Severity, and Recommendation."
* "Give me a short executive summary first, then the detailed findings."
* "List only the top five issues and include cell references."
This is especially helpful when you need to share the result with a reviewer or compare multiple outputs.
## Retry or edit messages
If Agent misunderstood your request, retry or edit your message with clearer instructions.
For example, change:
> Analyze this sheet.
To:
> Analyze the active sheet. Focus on formula consistency, hardcoded assumptions, and external links. Return a short list of issues with cell references.
## Use checklists for longer tasks
For multi-step work, Agent may show a checklist so you can follow progress. This is useful for workbook review, data cleanup, workbook comparison, or multi-sheet analysis.
If the checklist misses a step, tell Agent before it continues. If a step might change the workbook, ask Agent to preview the change or wait for approval.
## Manage long conversations
If a conversation gets long, Agent may summarize earlier messages so it can keep working. If Agent misses an important detail later, restate it in your next message.
Start a new conversation when you switch to a different workbook, topic, or review goal. This keeps the context cleaner and makes the conversation easier to find later.
## Tips
* Select the relevant range before asking about specific cells.
* Include sheet names and cell references when you know them.
* Ask for a preview before large edits.
* Ask Agent to explain assumptions before building new analysis.
* Use Skills for repeated instructions or team-specific conventions.
* Review important AI responses before relying on them. See [AI Guidelines](/assistant/agent/ai-guidelines.md).
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter:
```
GET https://docs.coherent.global/assistant/agent/chatting-with-agent.md?ask=
```
The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.