MCP
One line of client configuration puts eight HumanOps tools into every agent you run. It is the same service as the REST API through a second door: the same keys, the same limits, the same internal functions, and tool schemas generated from the same catalogue rows.
The MCP server is the same service through a second door: every tool calls the identical internal function the REST endpoint calls, and the tool input schemas are generated from the same catalogue rows. Paste one line into a client config and every agent you run gains these tools.
Endpoint https://quiescence.eu/humanops/mcp
Transport streamable HTTP, JSON-RPC 2.0
Auth Authorization: Bearer ho_live_… (or ho_test_… for the sandbox)
Versions 2025-06-18, 2025-03-26, 2024-11-05| Tool | What it does |
|---|---|
list_human_capabilities | What can be delegated to a person |
check_human_coverage | Can this be done in this place, and how fast |
quote_human_task | Exact price and deadline, before you commit |
create_human_task | Order the task (spends real money) |
get_human_task | Where a task has got to |
get_human_task_result | The answer, its confidence and its evidence |
cancel_human_task | Cancel and get the money back |
list_human_tasks | Every task this agent has ordered |
Copy-paste configuration for Claude Code, Claude Desktop, Cursor, VS Code, Windsurf, Cline, Goose and n8n — plus LangChain, the OpenAI Agents SDK and LlamaIndex — is on the install page.
The one thing to get right
create_human_task returns immediately with a reference. It does not wait, and no tool call ever will — a person then takes between an hour and a day. Report the reference, end the turn, and read get_human_task_result later.