read_doc
Reads one docs page as Markdown by the id search_docs gave, in parts small enough for a model's context.
Details #
read_doc gives back one page of the bundled docs as Markdown: the summary, the details, the parameters, the examples as JSON, and the gotchas. Links to other pages name their id in backticks, so the agent can read on.
Pages are at most 12,000 characters at a time. A longer page answers truncated: true and a nextOffset; pass it back as offset to read the next part.
{ "jsonrpc": "2.0", "id": 6, "method": "tools/call",
"params": { "name": "read_doc", "arguments": { "id": "step/typeText" } } }The answer is {id, title, kind, text, offset, totalChars, truncated, nextOffset?}. A bare id such as typeText works when only one page has it. Each page is also a resource, switchboard://docs/<id>.