Trusted by engineering teams at
Turn your API endpoints
into MCP tools any
AI agent can call.
Install the OpenTest MCP in your IDE. Test your endpoints from your agent as you build them, then publish them as a hosted MCP server with per-user auth and API keys — so any MCP-compatible agent, in any IDE, can call your API.
$ npx @opentest/mcp
The shift
APIs stop being something you document for humans and start being something AI agents natively consume.
Most AI coding agents are blind to the APIs you're building. They can read the code, but they can't call it. Testing means jumping out of your editor into Postman, copying URLs, hand-rolling auth headers, and pasting responses back into chat.
OpenTest closes that loop.
Now your agent can test endpoints, generate edge-case suites, detect schema drift, and run regression flows without leaving the editor. Publish your MCP so other teams — or your customers' agents — can call your API the same way.
How it works
Three steps between your repo and an agent-callable API.
Install in your IDE
Drop OpenTest into your MCP config once. No repo scan, no build step, no framework lock-in — it runs wherever your agent runs.
Test endpoints from your agent
Your agent calls your own API while you build — inspects responses, generates edge cases, runs regression flows, all inline with the code.
Publish a hosted MCP
One click turns your endpoints into a hosted MCP server with a URL you share. We host it; you stay in control of auth and keys.
One line of config
Plugs straight into your IDE.
Drop this into your MCP config. Your agent discovers every tool on startup — with full type information and auth pre-wired.
- Works with Cursor, Claude Code, Codex
- Typed inputs & validated responses
- Bring your own auth, or let us handle it
{
"mcpServers": {
"opentest": {
"command": "npx",
"args": ["-y", "@opentest/mcp"]
}
}
}From zero to agent-callable
Your APIs, instantly callable by Cursor and Claude Code.
Install OpenTest in your IDE
Add one block to your MCP config. Works with Claude Code, Cursor, Codex, Windsurf, Cline, Zed — any MCP-compatible agent.
Test your endpoints in-editor
Your agent hits your running API directly. Send requests, inspect responses, generate edge cases — without leaving the file you're writing.
Turn endpoints into MCP tools
Pick which endpoints should be agent-callable. OpenTest wraps each one as a typed tool with validation and auth, right inside your IDE.
We host it, you keep control
Ship a hosted MCP at your own URL. Issue per-user API keys, scope what each user can call, and revoke anytime. Your auth, your rules.
Built for
Teams shipping AI-native APIs.
Internal APIs
Let your team's agents call internal services without waiting for someone to hand-roll a bespoke integration.
Public APIs
Ship an MCP alongside your SDK. Your users' agents become first-class consumers of your product.
In-editor workflows
Stop jumping out of your editor to call an endpoint. Your agent calls it for you — typed, authed, and inline with the code you're writing.
Pricing