Documentation
Documentation
Mockd puts one observed endpoint between your agent and the tools it calls. These pages explain what happens behind it, in seven topics, each readable on its own.
The fastest path to a first trace
No account, no API key. Open the demo, paste the config into the MCP client you already use, and run your agent. The trace fills in as it calls tools.
claude_desktop_config.jsonjson
{
"mcpServers": {
"mockd-hubspot": {
"url": "https://mock.mock-d.com/s/<session-token>/mcp"
}
}
}The URL carries your session token. It's the only credential involved. There's no header and no key alongside it.
Topics
7- Core conceptsThe six words that come up everywhere, and what Mockd can and can't see.
- ConnectingOne line of config, no SDK, no key. What actually goes over the wire.
- Sessions & runsHow long a session lives, what a reset does, and which caps apply.
- Mock serversWhy the tool descriptions are real, and why tool calls affect each other.
- Injecting failuresRate limit, timeout, empty result, latency. Why "empty" matters most.
- Reading a traceWhat gets recorded per tool call, live and afterwards.
- Limits & dataWhat's synthetic, what's stored, and what doesn't exist yet.