What we do
MCP server development
Tool APIs agents can call.
Model Context Protocol servers for commerce — the JSON-RPC substrate that lets AI agents read your catalog and act on your store. Ours is public; the API is the demo.
- 1
Typed tools, not prompts — catalog search, product detail, stock, and cart actions exposed as JSON-RPC methods agents can actually rely on.
- 2
Public by default — our own endpoint answers tools/list from any agent, today. The API is the demo.
- 3
Operated, not abandoned — documented at /developers, specced in /openapi.json, rate-limited and versioned like a real API.
POST https://tendrils.io/mcp — call it yourself
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/list"
}
→ 200 OK
→ 11 tools: catalog, inventory,
configurator, contact, docs…
→ RateLimit-Remaining: 19Common questions
Asked before the first call.
- What does an MCP server do for a store?
- Exposes catalog search, product detail, stock, pricing and cart actions as tools agents call over JSON-RPC — agents transact instead of scrape.
- Can any AI agent use it?
- Any MCP-speaking client — our public endpoint at tendrils.io/mcp answers tools/list right now. Try it from your own agent.
- How is this different from a chatbot?
- A chatbot talks. A tool server acts — typed methods and structured data. The chatbot is one UI; the tool server is the substrate.
- Who maintains it after launch?
- We do. Versioned, documented at /developers, specced in /openapi.json — operated like any production API.
Ready for agents to transact on your store?
Talk to us