TasteGraph: The Taste Layer for AI Shopping Agents
TasteGraph builds what it calls taste technology: the judgment a good personal shopper brings, packaged so AI agents, retailers and publishers can use it. An AI assistant that’s helping someone shop can send TasteGraph everything it knows about the person, such as the occasion, their wardrobe, sizes, budget and style, and get back products that fit, with reasons, from a catalog of hundreds of thousands of items from premium retailers. TasteGraph also runs its own consumer app, Co-Shopper, a personal shopping assistant that learns your taste from your Gmail and Google Calendar.
The quick answer: TasteGraph is betting that as shopping moves into conversations with AI, the hard part shifts from finding products to choosing well for one person. Its approach treats a person’s taste as structured, dependent on the occasion and changing over time, rather than a list of keywords. It’s early, with a research-heavy team and little outside coverage so far. A disclosure: we make Carly, an AI assistant with its own memory that also works from your email and calendar, so we overlap on the personal side. Everything below comes from TasteGraph’s and Co-Shopper’s sites.
TasteGraph at a glance
Checked against TasteGraph’s site, research notes and privacy notice, and Co-Shopper’s site, on November 12, 2026.
| TasteGraph | |
|---|---|
| What it is | A taste and product-selection layer for commerce, built for AI agents |
| Products | Taste for Agents, Taste for Retailers, Taste for Publishers, and the Co-Shopper app |
| Categories | Fashion, jewelry and gadgets |
| Catalog | Hundreds of thousands of products from premium retailers |
| Consumer app | Co-Shopper, which learns your taste from Gmail and Google Calendar |
| Pricing | Not published; demos by request |
| Company | TasteGraph AI Inc.; co-founded by Clea O’Hana, Lia Kislev and Stéphane Benayoun |
The problem TasteGraph is solving
TasteGraph’s homepage starts with a blunt diagnosis: “Most of what agents find is AI slop.” The reason, it argues, is that “Agents know the person, but search with tools built for keywords. Rich context becomes basic filters.” An assistant might know that you need outfits for client meetings and a team dinner, that you already own black flats and an ivory blazer, and that you want to look “polished, never overdressed,” but a product search flattens all of that into a category and a size.
TasteGraph’s about page puts the shift this way: the hard part “is no longer finding products. It is choosing well for one particular person: what suits them, what works with what they already own, what is worth waiting for. That takes taste.”
How Taste for Agents works
Taste for Agents is the core product, “the taste layer for AI agents.” The flow, as TasteGraph describes it:
- Context in. The agent sends the full story: occasion, taste, wardrobe, sizes and budget, whether that’s “one line or several pages.”
- A picture of the need. TasteGraph turns the request into structured signals, such as dress code, constraints, pieces the person already owns and taste.
- A catalog built for judgment. It draws on products matched across retailers, checked for price and size, and described “in a stylist’s terms.”
- A model of the person. A model of their taste, informed by similar profiles, weighs each piece against the person and against the other pieces.
- An outfit, not a list. It returns an anchor piece and what completes it for the occasion and wardrobe, with alternates, live offers and why each fits. If the brief is too thin, it asks one clarifying question.
The homepage shows what that conversation looks like: TasteGraph asking the client’s agent whether the person likes heels (“Low heels only. Comfort comes first.”), whether they’ll be walking between meetings, and whether they prefer a loose or fitted cut.
Built for agents, not browsers
TasteGraph designed the product around how agents work. A person sends their AI agent one message, and “the agent requests access in its own name and receives its credentials at its own address.” The API is versioned, with machine-readable documentation and a public changelog, and breaking changes are announced 30 days ahead.
TasteGraph also makes a point about neutrality: “Every response is data the agent weighs with its own judgment. Nothing steers the agent, favors a retailer or speaks for its user.” For a service sitting between shoppers and stores, that’s an important promise to watch.
Taste for retailers and publishers
TasteGraph sells the same judgment to two other kinds of customer:
- Retailers can learn each customer’s taste from “what customers choose, keep, and respond to,” then select products from their own catalog and suggest how to pair them.
- Publishers and creators can have their own taste modeled, so their audience can “explore that perspective, ask what to buy, and discover how to put it together.”
Co-Shopper
TasteGraph’s own consumer app, Co-Shopper, shows the technology working for one person. “It reads your Gmail and Google Calendar to learn your taste, then recommends the pieces that are right for you.” You can ask it questions like “Should I get these?” and its site shows it scanning an inbox for sale and early-access emails from designer brands. TasteGraph’s homepage describes Co-Shopper as helping decide “what to buy, what to wait for, and what to pass on.” Co-Shopper has its own privacy notice, separate from TasteGraph’s website.
The research behind it
TasteGraph publishes research notes on what a model of taste needs to represent, keeping established research separate from its own position. Their titles summarize the thinking: “Taste Is Not a Vector,” “Preference Has a History,” “Context Is Part of the Preference,” “Similarity Is Not Understanding” and “Personal AI Needs a Theory of the Person.” That last note argues that “Personal AI is usually framed as a memory problem. The harder problem is deciding what accumulated evidence should mean about the person it describes.”
The company
TasteGraph AI Inc. was co-founded by Clea O’Hana, its CEO, Lia Kislev, its chief product officer, and Stéphane Benayoun, its CTO. The company says the team combines “more than a decade in e-commerce, shopping and human preference” with a background in AI research and large-scale AI systems. We found no published funding or independent press coverage, and TasteGraph doesn’t publish pricing; businesses request a demo.
How TasteGraph compares
TasteGraph belongs to the personalization side of agent infrastructure, next to memory layers like Supermemory, Honcho and Zep and the camera-roll SDK Rollup. Where those store and recall what an agent knows about someone in general, TasteGraph specializes in turning that knowledge into shopping decisions, with its own product catalog behind it.
If you want an assistant that knows you from your email and calendar for everything, not just shopping, Carly has memory built in: it saves named memory documents it reads and searches in later conversations, searches a knowledge base of files you upload by meaning, and keeps your to-dos, data tables and CRM in one workspace. It also labels and sorts email across all your Gmail and Outlook accounts by your rules, handles advanced scheduling, and makes and answers calls and texts, from inside ChatGPT or Claude or by text.
Who TasteGraph suits
TasteGraph is a good fit if:
- You’re building an AI assistant that helps people shop and want recommendations that reflect real taste.
- You’re a fashion, jewelry or gadget retailer that wants personalized selection and styling.
- You’re a publisher or creator who wants your audience to shop through your point of view.
Look elsewhere if:
- You need categories beyond fashion, jewelry and gadgets.
- You want published pricing and an established track record.
- You want general memory for an assistant rather than shopping judgment.
FAQ
What is TasteGraph?
TasteGraph is a company that builds taste technology for commerce. Its Taste for Agents API turns what an AI agent knows about a person into product selections and outfits that fit, from a catalog of products from premium retailers.
What is Co-Shopper?
Co-Shopper is TasteGraph’s personal shopping assistant. It reads your Gmail and Google Calendar to learn your taste and recommends pieces that suit you, including what to buy, wait for or skip.
Who is TasteGraph for?
AI agents and the apps that build them, fashion and other retailers, and publishers and creators who want their audience to shop through their taste.
How much does TasteGraph cost?
TasteGraph doesn’t publish pricing. Businesses request a demo.
Who founded TasteGraph?
Clea O’Hana, Lia Kislev and Stéphane Benayoun co-founded TasteGraph AI Inc.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
