Palisade: Product, team and how it was built
Palisade (AI sales agents for online marketplaces, one agent per visitor, United States): Live in Y Combinator's Summer 2026 batch, November 2026. Evidence grade E: no revenue figure was checked. Team: 1 founder; Y Combinator lists a team size of 1 in November 2026. Launched 14 August 2026; still running.Added 7 November 2026
- Outcome
- Live in Y Combinator's Summer 2026 batch, November 2026
- Evidence
- E no revenue figure was checked
- Launch
- 14 August 2026
- Status
- Still running
- Team
- 1 founder; Y Combinator lists a team size of 1 in November 2026
- Outside funding
- Y Combinator, Summer 2026 batch; the amount is not stated in the sources read
- Main channel
- Not stated
What is not verified
No revenue figure was found, and no price. The only figure, over US$2.7M in GMV since the start of the batch, is one statement by the founder in the Y Combinator launch post of 14 August 2026; it is the value of the customers' sales, not Palisade's income, and the post names no customer. Looked for: a payment-processor profile, a founder interview with figures, press coverage. A reader could look for a named customer case study or a later interview. This story is grade E: the business is real and its product, price and customers are described from public pages, but no revenue figure was checked. Aiolum shows it as an idea worth studying, not as proof that it makes money.
Palisade sells AI sales agents, programs built on a language model that talk with a shopper and take actions on the shop's site for them: open listings, add to cart, book. Each visitor to a marketplace (a site where many sellers or owners list goods or rentals) gets an agent of its own that keeps a memory of the visitor. Owners and sellers get an agent that explains their own data, and the marketplace staff get leads in their CRM (customer relationship management software, the database of customers and contacts).
Jonathan Salama, a Princeton computer science graduate who worked as a software engineer at Amazon and as a founding engineer at Sage AI Labs, started it alone in 2026. It launched on Y Combinator's site on 14 August 2026 in the Summer 2026 batch. The founder wrote in that post that the agents had driven over US$2.7M in GMV (gross merchandise value, the total value of what customers bought through the agents) since the start of the batch.
No revenue figure was found, so this page is grade E. The US$2.7M that the founder stated is the value of the marketplaces' sales, not Palisade's income, and no source says what Palisade charges.

How much revenue does Palisade make?
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Status, November 2026 | live; Y Combinator lists the company as Active | C | Y Combinator company page and the Palisade home page, both read 7 November 2026. The home page says "Live in production today". | [1, 3] |
| Founded | 2026, San Francisco | C | Y Combinator company page. The privacy policy names the company behind Palisade as 29 Labs, Inc., a Delaware corporation headquartered in San Francisco. | [1, 5] |
| Public launch | 14 August 2026 | C | Y Combinator launch post, "Palisade: Personalized sales agents for online marketplaces", with 84 votes when read on 7 November 2026. | [2] |
| Customer sales through the agents (GMV), August 2026 | over US$2.7M since the start of the batch | D | One statement by the founder in the launch post of 14 August 2026. GMV is the value of what the marketplaces' customers bought, not Palisade's revenue. The post names no customer and no period length. | [2] |
| Team | 1 person, the founder | C | Y Combinator lists a team size of 1. Extruct, a company directory, gives "1-10" employees. | [1, 6] |
| Funding | Y Combinator, Summer 2026 batch | C | Y Combinator company page. No source read here states the amount invested. | [1] |
| Prices | not published | C | Demo only. The terms of 11 June 2026 describe paid plans billed monthly or yearly, with no amounts. | [3, 4] |
| Integration | one script tag; Palisade connects the booking engine, property management system, CRM and catalog itself | C | Home page: "We wire in the deep integrations ourselves". | [3] |
| Channels to the shopper | the website, SMS and email | C | Home page and launch post. Agents set "standing watchers" for price drops and new stock and then text or email the shopper. | [2, 3] |
| Earliest archived product script | 15 May 2026 | C | The Internet Archive holds a capture of a chat widget script on an api.palisade.run subdomain from 15 May 2026, and scripts named assistant, capture and replay on cdn.palisade.run from 25 August 2026. | [7] |
| Time from launch to the headline | 3 months | calc | Calculated: 14 August 2026 to 7 November 2026, 85 days. | [1, 2] |
How to read this table: a grade C value is a statement by the company on its own site or its Y Combinator pages, or a directory that repeats them. The grade D row is one founder statement on one date. A "calc" value was computed on this page. Nothing here was checked by a payment processor, a filing or a journalist.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Contract after a demo | no prices published |
The site has no pricing page. Every page leads to "Get a demo" or a contact form, and the launch post asks marketplaces to email the founder. The company's terms of 11 June 2026 say that fees of a paid plan are "billed in advance on a monthly or annual basis" and are non-refundable, but they give no amounts.
When did Palisade launch, and how did it grow?
From an archived widget script in May 2026 to a Y Combinator launch in August 2026. Each entry carries its grade.
- 15 May 2026CThe Internet Archive captures a Palisade chat widget script, the earliest archived trace of the product. [7]
- 11 June 2026CTerms of service dated this day describe Palisade as "an adaptive user interface customization platform" with an SDK, a dashboard and APIs. [4]
- 14 August 2026DLaunch post in Y Combinator's Summer 2026 batch. The founder writes that the agents had driven over US$2.7M in GMV since the batch started. [2]
- 25 August 2026CScripts named assistant, capture and replay are archived on the company's content delivery address, cdn.palisade.run. [7]
- November 2026CThe site sells "agents that make markets" by demo. Y Combinator lists the company as Active with 1 person. [1, 3]
Why did Palisade work?
Four causes, in the order they mattered. Each one is a reusable lesson.
Domain experience
The founder built software at Amazon that helped influencers sell products, then worked on early iMessage shopping agents at Sage AI Labs. Palisade applies that work to marketplaces with expensive, considered purchases: vacation rentals, yachts, collectibles.
An agent that acts, not only answers
A support chatbot waits for a question. Palisade's agent reads the visitor's behavior, opens the conversation when it judges the moment right, navigates the site for the visitor and finishes the checkout. It keeps a memory of each visitor and follows up by text or email when a price drops or new stock arrives.
Both sides of the market
The same install gives sellers and owners an agent that explains their occupancy or sales in plain English and proposes listing changes that a person approves, and gives the marketplace staff leads with the full session attached.
How it was built
These details come from the company's own site, its terms and its Y Combinator launch post. They are grade C. [2, 3, 4, 5, 7]
- Customers add one script tag; the home page shows it loaded with Next.js from cdn.palisade.run.
- Palisade connects each customer's booking engine, property management system, CRM and catalog on its own side.
- Archived script names on cdn.palisade.run suggest separate parts for the assistant, lead capture and session replay.
- Outbound text messages run through an SMS delivery provider that the policies do not name; the company's SMS page describes the consent forms.
- No source names the language model. The launch post compares each agent with "an omnipotent Claude embedded into the platform".
Who tried this before Palisade, and why did they fail?
Three companies in our database of failed startups promised a personal salesperson for online shoppers and failed. The values below are from that database and are not independently verified.
Operator
The promiseA personal shopper in your pocket, mixing human staff and software to find hard-to-find items.
Why it diedHuman staff made each order expensive, and Amazon and Walmart were faster and cheaper without a personal touch.
RightHello
The promiseLive chat for online shops that started conversations with visitors at the right moment to turn them into buyers.
Why it diedIntercom raised far more money and grew into a full suite, and RightHello could not afford to compete.
Vanti
The promiseConversational AI that guided shoppers through product discovery, recommendations and purchase.
Why it diedIt could not show that chatbot shopping beat a normal online shop on conversion, and integration was costly.
The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.
What changed
Operator needed people to do the selling, and RightHello and Vanti needed scripted chat flows that could not act on the shop's site. A language-model agent can read a whole catalog, operate the site's own buttons and remember each shopper, at a cost per visitor that a person cannot match.
Vanti's open question is still open: whether an agent sells more than a well-built search page. Palisade's only answer so far is the founder's own customer-sales figure, with no conversion study published.
Can one person copy Palisade now?
Hard for one person.
Each customer needs its booking engine, CRM and catalog connected, which the company does itself, and the selling is by demo to marketplaces. The idea can be copied in a narrower form: one agent for one type of marketplace software that many small operators share, so that one integration serves many customers.
Where it is still open: Marketplaces for used cars, wedding venues, private tutors and second-hand luxury goods, where a human salesperson is normal offline.
- 1Pick a kind of marketplace where purchases are large and buyers compare for days: rentals, boats, collectibles.
- 2Connect to the inventory and booking system that most operators of that kind use.
- 3Let the agent act on the site, keep a memory of each visitor and follow up only with the visitor's consent.
- 4Measure the sales the agent closes, and publish how they were counted.
What could go wrong for Palisade?
The risks are the mirror image of the causes above.
What is not verified
No revenue figure was found, and no price. The only figure, over US$2.7M in GMV since the start of the batch, is one statement by the founder in the Y Combinator launch post of 14 August 2026; it is the value of the customers' sales, not Palisade's income, and the post names no customer. Looked for: a payment-processor profile, a founder interview with figures, press coverage. A reader could look for a named customer case study or a later interview.
Does the shopper know it is an AI?
The agents chat on the customer's site and send texts and emails to shoppers. The home page lists "As visible or as invisible as your brand wants" among the features, and also "Human handoff built in". No page read says whether the agent tells the shopper that it is an AI, or how consent for its texts to shoppers is collected; the SMS consent page covers only Palisade's own messages.
Demo names under "Live in production today"
The home page shows three brands, Shorebreak, Vaulted and Atelier, under the words "Live in production today", in a browser frame with the address shorebreak.palisade.run. That address did not open a web page on 7 November 2026, and no source ties these names to paying customers.
Tracking of visitors
The product reads each visitor's behavior to judge intent, keeps a memory of searches and near-purchases, and an archived script is named replay. Palisade's privacy policy covers its own site and services; no page read says what the marketplaces must tell their own visitors about this tracking.
Competition and model providers
The site already has Alvas AI, a Shopify popup that answers shopper questions; Palisade goes further by acting on the site and following up by text. No source names the language model behind the agents; the launch post's comparison with Claude suggests, but does not show, a dependence on Anthropic or another model provider.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- C Y Combinator company page: Palisade, AI-native sales agents that run marketplacesRead in full, 7 November 2026, from the page's embedded data. Company-written text. Source for the founding year, team size, batch, status and founder biography.
- C Y Combinator launch post, 14 August 2026: Palisade, Personalized sales agents for online marketplacesRead in full. The founder's post. Source for the GMV statement, the product description and the backstory.
- C Palisade home page, read 7 November 2026Read in full. Source for the features, the integration, the demo names and the absence of prices.
- C Palisade terms and conditions, last updated 11 June 2026Read in full. Source for the payment terms and the earlier product description.
- C Palisade privacy policy and SMS updates page, read 7 November 2026Read both in full; the SMS page is at /sms-opt-in. Source for the company name 29 Labs, Inc. and the SMS consent forms.
- C Extruct company profile: PalisadeRead in full. A directory that repeats the Y Combinator data and gives 1-10 employees.
- C Internet Archive capture index for palisade.run, read 7 November 2026Read the list of captures. Source for the dates of the earliest archived product scripts. The scripts themselves were not opened.
- C Database of failed startups, local exportLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.
Run this company? Send corrections or a verified figure to aiolum@livyden.com