Capabilities is where you turn an agent's built-in functions on and off — the ready-made abilities the platform ships, like web search, running code, or creating a document.
Overview
The tab is described as "Choose the built-in functions this agent can use to respond and complete tasks." It opens on Choose what the agent can do, with the note "Enable the capabilities the agent can use when responding to requests and completing tasks." Unlike Tools — where you add your own connectors and actions — Capabilities is a settings list of everything that's built in: each one shows a toggle.
Two counts sit on the screen and mean different things: the pill beside the section title Capabilities is how many built-in capabilities exist in total, and the pill beside Choose what the agent can do is how many are currently switched on for this agent.
Search to find a capability and switch it on — it's saved for you. Some capabilities work with just the toggle. Many others — such as Web Search, Execute Code and the document creators — also show a Configure (sliders) button, available once the capability is switched on. It opens a Configure capability drawer with settings specific to that capability, such as which model powers it. The button is disabled while the capability is off, and turning a capability off doesn't erase whatever you configured — it's kept and reused if you switch the capability back on.
Turning a Capability On or Off
Find the capability in the list and use its toggle. Each toggle saves immediately to the agent (there's no separate Save button on this tab); if a save doesn't go through, the switch flips back. Turn on only what the agent needs: extra capabilities widen what the model may attempt, so a focused set behaves more predictably.
The list loads about 30 capabilities at a time and fetches more as you scroll, and the search box queries the whole catalog rather than only the rows already on screen — so searching is the reliable way to find something, not scrolling.
Why the List Might Look Empty
If the catalog can't be loaded — or genuinely has nothing in it — the tab reads No tools available, "There are no standard tools to add right now." That same neutral message covers a failed load, so if you were expecting capabilities here, reload the section before assuming your workspace has none.
A search that matches nothing is different: that shows No Results Found with a Clear Search button, and clearing the search brings the full list back.
Built-in Capabilities
The tab shows one searchable list. Which capabilities are available can change as the platform adds more; the list on the tab is always the current set. At the time of writing it includes:
Capability | Description | Has Configure settings |
|---|---|---|
Analyse File | Reads an uploaded attachment and adds its content to the conversation context. | — |
Build App | Generates an interactive HTML artifact (dashboards, visualizations, mini-apps and interactive reports) rendered in the canvas beside the chat. | Yes |
Create DOCX / PDF | Generates a Microsoft Word document or a PDF. | Yes |
Create Docx/ PDF with template | Generates a Word document or PDF by filling one of your uploaded .docx templates, keeping the template's styling, tables and layout. | Yes |
Create Excel / CSV | Generates a Microsoft Excel spreadsheet. | Yes |
Create Markdown | Produces a downloadable .md or .json file when the file itself is the deliverable, such as reports, notes, READMEs, structured data and config payloads. | Yes |
Create PPTX | Generates a Microsoft PowerPoint presentation. | Yes |
Create PPTX with Template | Generates a PowerPoint from an uploaded .pptx template. | Yes |
Deep Research | Runs multi-step research workflows that synthesize across many sources. | Yes |
Execute Code | Runs bash or Python and returns the execution results. | Yes |
Generate Chart | Creates charts and graphs from provided data. | Yes |
Generate Image | Creates images from text prompts or instructions. | Yes |
Ingest Knowledge | Imports and stores new knowledge into the system. | — |
Planning Tool | Manages and tracks a list of tasks to complete. | — |
Q&A on Attachments | Adds the content of uploaded attachments to the conversation context so questions can be answered against them. | — |
Render Artifact | Opens a generated file or HTML/Markdown output in the interactive canvas beside the chat: paged preview for PDF, DOCX and PPTX, sheet preview for XLSX, live render for HTML and Markdown. | Yes |
Request Clarification | Requests additional details to resolve ambiguity. | — |
Web Search | Searches the live web for current information and recent events. | Yes |
Capabilities vs Tools vs Skills
Concept | What it is |
|---|---|
Capabilities | Built-in functions you just switch on — no connection to set up, and often no configuration beyond the toggle (many, like Web Search, offer a Configure drawer for capability-specific settings once they're on). |
Tools | Things you add and set up: connector actions, MCP servers, custom HTTP endpoints, and workflows-as-tools that reach into your systems. |
Skills | Reusable expertise you attach to shape how the agent approaches a kind of work. |
An agent can use all three at once. If you're looking for a built-in ability and can't find it under Tools, check Capabilities — the built-in catalog lives here now.