The agent marketplace formachine-payable compute.
Sella turns curated datasets, evals, live signals, workflows, and APIs into products agents can discover, preview, pay for, and use on their own.
The asset marketplace for autonomous agents starts with trusted data.
Sella begins with curated datasets, evals, and paid retrieval, then expands into live signals, workflows, connectors, and models that agents can buy when a task demands it.
Everything needed to turn data into a queryable, payable product.
Sella handles discovery, access control, and payment settlement. Publishers control what agents can access and at what price.
List a dataset once. Any MCP-compatible agent — Claude, ChatGPT, Cursor, or a custom worker — can find and query it without a custom integration.
Agents pay only when they access data. Payments settle on-chain in USDC — no invoices, no escrow, no minimum payout thresholds.
Publish a free preview for qualification. Lock the full dataset behind a price. Evaluation and monetisation stay separate.
Manage API keys, dataset visibility, pricing, and payout settings from a single operator view.
A storefront built for autonomous buyers.
Agents filter by category, tier, and recency. Humans get the same interface. Every dataset page surfaces structured metadata so buyers — human or machine — can assess before they commit.
Browse datasetsOne endpoint. Any agent stack.
The MCP surface exposes search, preview, and paid access in a format agent clients already understand. No custom adapter. No SDK required.
Open docsShip data. Watch what agents actually buy.
Keep premium data locked, offer free samples for qualification, and track access patterns across your catalogue from one dashboard.
View dashboardWire Sella into any agent in two lines of config.
The MCP endpoint exposes listing, search, trial, and retrieval in a format agent clients already understand. No custom adapter layer required.
- 1.
Generate an API key from your publisher dashboard, or complete OAuth if your client supports it.
- 2.
Point your agent at the MCP endpoint. Claude Desktop, ChatGPT, Cursor, and custom workers all use the same config.
- 3.
Agents can trial exposed previews first, then premium rows surface payment instructions when they buy.
1import fetch from "node-fetch";
2
3const response = await fetch("https://sella.dev/api/mcp", {
4 method: "POST",
5 headers: {
6 "Content-Type": "application/json",
7 Authorization: "Bearer YOUR_API_KEY",
8 },
9 body: JSON.stringify({
10 jsonrpc: "2.0",
11 id: 1,
12 method: "tools/call",
13 params: {
14 name: "search_datasets",
15 arguments: {
16 query: "market microstructure",
17 },
18 },
19 }),
20});
21
22console.log(await response.json());Built for the workflows agents already run.
Track niche pricing datasets, expose a free preview, and let trading agents buy the premium slice only when the signal is relevant.
