> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.rightbrain.ai/v-1/docs/patterns/features/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.rightbrain.ai/_mcp/server.

# AI Features Integration

> Build intelligent product features with Rightbrain Tasks that your customers interact with directly

Embed AI capabilities directly into your product to transform user experiences. Rightbrain [tasks](/docs/concepts/tasks) make it simple to add intelligent features your customers interact with — a single task for a focused capability, or an [agent](/docs/concepts/agents) when the feature needs to reason across several steps.

## **The Pattern**:

#### Trigger

User triggers an action in your product (e.g. clicks “Generate”, uploads a file, submits a form)

#### Receive

Your app captures the user’s input and any contextual data and sends these to a **Rightbrain Task** via API

#### Run

The Task processes the request using your defined prompt, schema, and model configuration.

#### Respond

A structured JSON result is returned to your app → you display it and continue the user workflow.

## Feature Categories

#### Content Generation

#### E-commerce Listings

**Input:** Product specs, images uploaded by seller

**Output:** SEO-optimised titles, descriptions, bullet points

**Typical user:** Marketplace sellers, e-commerce merchants

#### Job Postings

**Input:** Basic job details from employer

**Output:** Complete job description, requirements, benefits

**Typical user:** Companies posting jobs on your platform

#### Property Listings

* **Input:** Property details, photos from agent/owner
* **Output:** Compelling listing description, highlights, amenities
* **Typical user:** Real estate agents, property owners

#### Profile Bios

* **Input:** User’s background, experience, interests
* **Output:** Professional bio or profile summary
* **Customer:** Platform users, freelancers, professionals

#### Document Intelligence

#### Invoice Extraction

**Input:** Uploaded PDF invoices\
**Output:** Structured JSON with line items, totals, dates\
**Used in:** Accounting, ERP, expense management

#### Contract Summary

**Input:** Legal document text\
**Output:** Key clauses, parties, obligations\
**Used in:** Legaltech, procurement, compliance

> **Note**
>
> Attachments never touch your code as raw bytes. When an agent pulls a document — say a supplier invoice PDF from an inbound email — the fetching tool saves it once to a run-scoped file registry and hands the next task a `file_id`. A vision-capable task then reads the PDF natively, no separate extraction step, and returns structured fields.

#### Smart Search

#### Semantic Product Search

**Input:** Free-text query (“red running shoes under £100”)\
**Output:** Ranked product matches\
**Used in:** E-commerce, marketplaces

#### Knowledge Retrieval

**Input:** User question\
**Output:** Extracted answers from internal knowledge base\
**Used in:** Customer support, internal tools

#### Job matching

**Input:** CVs or portfolios\
**Output:** Structured profiles \
**Used in:** Recruitment, HR, Gig economy

#### Image Enhancement

#### Auto-Tagging

**Input:** Uploaded image\
**Output:** Alt text, SEO tags, accessibility metadata\
**Used in:** DAM systems, CMS platforms

#### Visual Search

**Input:** Photo of item\
**Output:** Similar product suggestions\
**Used in:** Fashion, furniture, art platforms

#### Image refinement

**Input:** Photos of products
**Output:** Refined image for product listing
**Used in:** Marketplaces, Retailer catalogues

## Example: E-commerce Product Listing Assistant

**The Feature**: Help sellers create compelling product listings

**User Flow**:

1. Seller uploads product photos and enters basic info (category, price)
2. Clicks "Generate Listing"
3. Task analyses images and inputs, returns structured JSON with title, description, and bullet points
4. Seller reviews in split-screen editor
5. Edits any sections that need adjustment
6. Clicks "Publish" when satisfied

**Why It Works**:

* **Stateless by design** – Each generation runs in isolation, ensuring zero data leakage between sellers or listings.
* **Predictable shape** – The JSON schema constrains the fields and value types returned by a successful run. It does not guarantee factual correctness; use [evals](/docs/production/evals) for behavior quality.
* **Dynamic and contextual** – The Task adapts to each seller’s product context, generating copy that fits their category, tone, and brand style.
* **Composable** – The same Task powers the bulk upload tool, mobile app, and partner APIs - one definition, multiple use cases.
* **Production-ready** – Observability, rollback, and version control built in, so every listing is generated with traceability and confidence.

> **Note**
>
> Start with one high-impact feature. Get it working well. Measure the results. Then expand to more use cases. The best AI features feel like natural extensions of your product, not bolt-on tools.

## Next Steps

#### [Build Your First Feature](/docs/quickstart/create-a-task)

Create a task and test it with your data

#### [Run tasks via API](/docs/api/run-tasks)

Technical integration details and examples

#### [Triggers & runs](/docs/concepts/triggers-and-runs)

Start work from product events and inspect each execution

#### [Evals](/docs/production/evals)

Test behavior before activating a new revision

> Build reliable AI agents that run inside your existing tools and workflows. Rightbrain developer documentation.