> For the complete documentation index, see [llms.txt](https://docs.manyfast.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.manyfast.io/en/use/ai-chat.md).

# AI Agent

Manyfast's AI goes beyond simple Q\&A — it's a powerful partner that directly edits and executes your planning. Connected to all of your Project data (PRD, Specifications, User Flow), it **understands instructions based on context and performs tasks automatically**.

### Using the AI Agent

#### 1. Core Capabilities

* Plan review and information gathering: Ask questions during planning and get fresh ideas.
* Direct item editing: Instructions like "Update the PRD overview" or "Add detailed policies to the payment Feature" let the AI directly modify your planning data.
* File upload analysis: Upload reference materials (PDF, Word, Excel, images, etc.) such as planning documents or meeting notes for the AI to analyze. (Up to 50MB per file)

#### 2. Two Work Modes

You can choose how the AI works by setting the mode in the bottom bar.

* Design mode: Before making changes, the AI works with you to refine the plan and reach agreement first.
* Execute mode: When you give an instruction, the AI immediately uses planning tools to modify data and produce results quickly.

#### 3. Choose from Various AI Models

Pick the optimal model based on your Project's nature and credits consumption.

* OpenAI (GPT-5.2, etc.): Excellent general-purpose performance, strong at idea generation.
* Google (Gemini 2.5 Pro, etc.): Optimized for long-context processing and research tasks.
* Anthropic (Claude series): Outstanding at writing detailed Specifications and following instructions.

***

### Chat Interaction and Execution Management

#### Starting a Chat

* New chat and history: Use the pencil button in the top right to start a new chat, and the clock button to view previous conversation history.
* Item tagging: Type '@' in the chat input to select a specific item from the list as context and give focused instructions about that item.

#### Reviewing and Applying AI Edits

When the AI edits content based on your instructions, you review the changes and decide whether to apply them.

* Edit notifications: When the AI makes changes, a blue arrow button appears in the tree view. Click it to preview the edits in the directory view.
* Approve and reject: You can approve or reject changes individually for only the items you want, or use the bottom bar to approve/reject all at once.

### Credits Consumption <a href="#ai-credit" id="ai-credit"></a>

AI chat and generation features consume credits. The amount consumed varies each time based on the AI model in use and the volume of information processed. You can check your remaining credits at the bottom of the chat input or on the [Credits](/en/start/plan.md#undefined-10) page.

* The more and the more complex the work, the more credits are deducted.
* Using more advanced models may deduct more credits.
* Credits are also deducted for responses that don't involve AI edits (simple chat).
* Even if you press the stop button, credits for information already processed will still be deducted.
* Every task consumes a minimum of 1 credit.
  * If the AI fails the task, the 1 credit consumed at the start is refunded.

#### Credits Consumption by AI Model

* Anthropic (Claude) models consume the most credits.
  * Even a conversation of typical complexity deducts 10–20 credits.
  * Requesting large-volume tasks can deduct 50 credits or more.
* Google (Gemini) and OpenAI (ChatGPT) consume similar amounts.
  * A conversation of typical complexity deducts 5–10 credits.


---

# 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.manyfast.io/en/use/ai-chat.md?ask=<question>
```

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.