← wickwatch home

Model Context Protocol

Ask your AI assistant whether your trading bots are running

wickwatch has a read-only Model Context Protocol (MCP) endpoint. An AI client that speaks MCP reads the same data as the dashboard and answers questions like “Are all bots running?”, “Why did this one stop last night?” or “How much of the daily loss is used?”.

What an assistant can read

What it cannot do

Anything that changes something: there are no tools to start, stop, configure or trade, on purpose. Log lines, order labels and comments are written by bots and brokers and reach the model as untrusted text. A line in a log that reads like an instruction must not be able to stop a bot or close a position.

Set it up

  1. In wickwatch, open the user menu → API tokens and create a token. The viewer role is enough.
  2. Add the address shown on that page to your client, with the header Authorization: Bearer <token>.

With Claude Code, for example:

claude mcp add --transport http wickwatch https://wickwatch.example.com/mcp \
  --header "Authorization: Bearer ww_…"

Clients configured with JSON take the same address and header. The example, all tools with their arguments and troubleshooting are in docs/MCP.md.

Security