Documentation

The API answers HTTP requests in JSON. Reads are GET requests and writes are POST requests.

The MCP server offers the same procedures as tools to an MCP client such as Claude, ChatGPT or Claude Code. A tool is named by its path with an underscore between the segments, so /businessSimulation/list is the tool businessSimulation_list.

Each agent credential is rate limited to a burst of 30 requests, then 1 per second, and a request past that is answered with 429 and no body, as is a credential that is not accepted. An error from a procedure comes as JSON with the fields code, status, message and data.

Create an agent credential on your account page. It is shown only once and goes in the Authorization header as a bearer token.

An MCP client can connect through OAuth instead. You sign in and allow the connection, and the credential is created for you.

claude mcp add --transport http techtropy https://api.techtropy.com/mcp --header "Authorization: Bearer <secret>"

An agent credential acts in one organization as the person who created it and can do no more than that person can there. It reads, or reads and writes. An organization's administrators decide whether credentials are allowed in it, and the instructor decides for each simulation whether agents may play.

Each procedure with its inputs and what it returns, grouped by area. The OpenAPI specification holds the responses in full and loads into your own tools.