An agent that other agents can work with.
Releva is the sales agent inside a store. Your assistant, your app or your own agent can ask it questions over MCP, and feed it, read it and trigger it through the APIs.
- An MCP server for assistants such as Claude and ChatGPT
- Events, search, subscribers, feeds and webhooks through the API
- Every page of this site published for assistants: llms.txt and llms-full.txt
Everything the agent knows about a shopper comes in as events, and everything it decides goes out as a placement, a message or an audience. Both ends are open. Your systems push events in and read decisions out through documented APIs and webhooks, and an assistant can ask the store's agent directly over MCP.
An assistant asks through the MCP server. Your systems push site, app and custom events, search events and subscriber sync, and pull product feeds filled per shopper, event history and webhooks on journey steps.
The Marketing co-pilot (in beta) is the first agent built on that memory. The MCP server gives the same memory to the assistants your team already uses.
What an agent can do through the API today
Send what happens
Page views, product views, carts, orders and your own custom events go in through the events API, from the browser, the app or your server. A search event API records what shoppers look for. The subscriber sync API keeps consent and contact details current. The documentation lists each endpoint and its fields.
Read what the agent decided
Product feeds come back filled for one shopper, with your own rules on top (boost, bury, exclude). Webhooks fire on journey steps, so your systems can act when the agent does. Event history is readable per profile.
Where it runs
Shopify and plug-in platforms go live the same day; other platforms and custom stores take 3 to 5 business days for all tools. Mobile SDKs cover iOS and Android. The full list is on platforms.
The MCP server, in production
The MCP server is in production. An assistant connected to the store's MCP server can ask in natural language: which products fit this shopper, what to recommend next to this basket, the status of an order, what a journey did last week, and it can trigger a message within the rules the store set. Every answer comes from the same memory the agent uses on the site and in the inbox, with the store's own catalog, stock and prices.
Access is per store, with a key the store issues, and every call is logged on the shopper's profile like any other interaction. To connect your assistant, [book 30 minutes](BOOK) or write to info@releva.ai and we send the endpoint and the tool list for your store.
This site, for assistants
Every page here is written answer-first, carries structured data, and is listed in llms.txt. The full text of every indexed page is in llms-full.txt. Our robots.txt allows the crawlers of OpenAI, Anthropic, Perplexity, Google and Apple. If you quote us, use the page URL and the date on the page.
Where to look
API documentation
Events, search events, subscriber sync, product feeds, webhooks and the B2B APIs.
Open the docs →ProductHow it decides
What the agent does with the events: your goal, per-person decisions, a holdout for proof.
See the method →FeaturesAll features
The full list, grouped by job, including the MCP server and the APIs.
See the list →
See which of last week's visitors you missed
Thirty minutes, with your store and ad accounts open. Then 30 days free. A holdout group decides: if the agent doesn't add orders in 30 days, you don't pay.
Book 30 minutes
You'll talk to one of us.