Here is a complete list of each tool utilized by each agent task in ChatGPT (as of GPT-4o / Pro environment) along with their exact names, how they are invoked, and parameter descriptions:
Each of these tools is modular and is invoked during specific types of user requests (e.g. file uploads, web browsing, math solving, etc.) via agent orchestration logic.
### 🧠 python — Code Interpreter / Advanced Data Analysis (ADA)
❯ Name in API: python
❯ Usage: Executes Python code in a sandboxed Jupyter environment.
❯ What it does:
- Data analysis, plots, math.
- File parsing: CSV, Excel, JSON, PDF, images.
- Statistical operations and visualizations.
- Custom algorithms (sorting, optimization, etc.)
❯ Parameters (used internally):
No external parameters exposed, but internally it:
- Receives code as string
- Executes and returns
stdout, stderr, result (if any)
### 🖼️ image_gen — DALLE Image Generation & Editing
❯ Name in API: image_gen
❯ What it does:
- Generates or edits images based on user prompts or uploaded references.
- Can modify existing images with object insertion, style transfer, cleanup, etc.
❯ Tools within image_gen:
| Function | Parameters | Description |
|---|
text2im | prompt: str size: str (e.g. '1024x1024') n: int transparent_background: bool referenced_image_ids: List[str] | Generates or edits images based on textual description. |
| Referenced image IDs are required for inpainting or transformations. | | |
### 📎 canmore — Code & Document Canvas Editor
❯ Name in API: canmore
❯ What it does:
- Used for live-editing long code or document files.
- Allows persistent editing and commenting in the right-hand canvas view.
❯ Tools within canmore:
| Function | Parameters | Description |
|---|
create_textdoc | { name: str, type: str, content: str } | Creates a new document or code file in the canvas. |
update_textdoc | { updates: [{ pattern: str, multiple: bool, replacement: str }] } | Applies regex-based updates to the canvas text. |
comment_textdoc | { comments: [{ pattern: str, comment: str }] } | Leaves inline comments in the canvas file. |
### 🌐 web — Browser / Web Search Tool
❯ Name in API: web
❯ What it does:
- Enables up-to-date, real-time search and data retrieval from the internet.
- Used when the model lacks knowledge about current events, product details, or localized info.
❯ Tools within web:
| Function | Parameters | Description |
|---|
search() | None (internally uses user query) | Issues a search engine query. |
open_url(url: str) | URL as string | Opens a specific web page and parses it. |
❯ What it does:
- Triggers parsing via
python or canmore or image_gen, depending on file type:
.csv, .xlsx → python.docx, .txt, .md → canmore.png, .jpg, .pdf → image_gen or python depending on task
| Task | Tool | Call |
|---|
| Math & Code | python | python |
| Image generation/editing | image_gen | text2im |
| File parsing / editing | canmore | create_textdoc, update_textdoc, comment_textdoc |
| Web browsing | web | search(), open_url(url: str) |
If you'd like, I can also export this list as JSON, YAML, or a markdown cheat sheet for dev usage.
