Personalised Recommendations Agent
Show every customer recommendations that actually fit — drawn live from their history and your catalogue. No generic “you might also like”. It reasons over real data and can even explain why.
The problem
Generic recommendations don't convert
“You might also like” blocks ignore where the customer actually is in their journey. They're rule-based guesses — and customers can tell. The result is decision fatigue and lost revenue on every session.
A real recommendation engine usually means an ML pipeline, training data, and a dedicated team. Most businesses can't justify that just to suggest the next product.
The solution
Recommendations that actually fit
A Recommendations Agent reasons over your live customer history and catalogue through MCP. It suggests genuinely relevant next steps — and can explain why, which builds trust and clicks.
No model training, no ML infrastructure. Configure it once, call it from your app or email, and your customer data stays in your environment.
Out of the box
Real questions, real answers.
Tuned to each customer.
The agent reasons over live history and your catalogue — so suggestions feel personal instead of templated.
“What should I buy next?”
Suggests a genuinely relevant next pick from your catalogue.
“Got anything like this?”
Finds close matches based on the item they're viewing.
“What goes with my last order?”
Recommends complementary items that actually fit.
“I'm new here — where do I start?”
Guides first-timers with context-aware suggestions.
“Show me something in my size.”
Filters recommendations by their saved preferences.
“Why are you suggesting this?”
Explains the reasoning to build trust and clicks.
How it works
Four steps in SentientOne.
Live in days, not months.
Create the agent
In Agents, add a new agent, pick a model (e.g. OpenAI · gpt-4o), and write a short system prompt that defines its job.
Name
Provider
Model
Temperature
System Prompt
Connect your Insights API
On the agent's MCP tab, add your Insights API as an MCP server. SentientOne discovers the get_recommendations tool — your credentials stay in your environment.
Name
Transport
URL
Auth Type
Test it in the Playground
Ask real questions in the Playground. The agent calls your API, reads the live response, and replies in plain language. Tweak the prompt until it's right.
Go live with one request
Grab your API key and send a single POST from your app. No AI SDK — anything that can make an HTTP request works.
Platform API key
API endpoint
POST /v1/chat
X-Api-Key: sk-live-•••
X-Agent-Id: personalised-recommendations-agent
{ "message": "What should I buy next?" }Why SentientOne
Why teams ship this with us.
Not a 6-month engineering project.
No AI team required
Skip the ML hires, the prompt infrastructure, and the model plumbing. Configure the agent in the dashboard and you're done — we'll even set up the MCP server for you at no extra cost.
Your data stays yours
The MCP server runs on your infrastructure. SentientOne only receives the tool response — never your raw database, credentials, or records. Self-host the whole platform if you need to.
Works with your stack
Connect any REST or gRPC API through MCP. One HTTP endpoint plugs into React, Flutter, Python, .NET, Go — anything that can make a request. No SDK lock-in.
Switch models, never rebuild
Run GPT-4o today, Claude tomorrow — change it from a dropdown. When your API changes, you update one tool definition. No retraining, no redeploys.
“We replaced our rule-based 'you might also like' with an agent that actually reasons over each customer. Revenue per session jumped, and we didn't hire a single ML engineer.”
The outcome
What you get
15–30%
typical lift in revenue per session
Day one
works on your existing data — no training
Private
your customer data stays in your environment
Questions
Common questions
How long does it take to go live?
Most teams are answering real questions within a few days. There's no model training and no AI pipeline to build — just connect your API and configure the agent.
Do I need AI engineers?
No. The agent is configured in the dashboard, and we'll set up the MCP server that wraps your Insights API for you at no extra cost.
Is my data safe?
Yes. The MCP server runs in your environment and only returns the specific tool response. Your database, credentials, and raw records never reach SentientOne.
What happens for questions it can't answer?
You define the boundaries. Anything outside product recommendations is handed off cleanly to your team with full context.
Which models can I use?
GPT-4o, Claude, Gemini, Groq and more. Pick one from a dropdown and switch any time — your setup and integration stay exactly the same.
How does it reach my users?
One HTTP endpoint. Drop it into your existing chat widget, app, or site — React, Flutter, Python, .NET, Go, anything that makes a request.
Free white papers
Go deeper on this
Improve Your Search with Agentic AI
Most businesses still run a keyword search bar over a catalogue that customers describe in plain language. The gap between the two is lost revenue — every day. This guide shows how to move from keyword matching to natural-language search, and how to deploy a production-grade Product Search Agent in days, not months.
Improve Your Customer Support with AI Agents
Support queues grow faster than headcount, and every minute a customer waits chips away at trust. Most teams paper over the gap with canned macros and after-hours auto-replies that never actually resolve the issue. This guide shows how AI agents handle the repetitive tickets end to end, escalate the ones that need a human with full context, and give your team back the hours they spend on copy-paste answers.
More patterns
Explore other use cases
Order Status Agent via MCP
Let customers ask about their orders in plain English.
Business Intelligence Reporting Agent
Turn your database into a conversation.
Customer Support Automation Agent
Handle 70% of tickets before they reach your team.
Product Discovery & Catalogue Agent
Let users search your catalogue by describing what they want.
Get started