+3

lib/claude.mli

··· 210 210 module Model = Model 211 211 (** Claude AI model identifiers. *) 212 212 213 213 + module Structured_output = Structured_output 214 214 + (** Structured output configuration using JSON Schema. *) 215 215 + 213 216 (** {1 Custom Tools (MCP)} 214 217 215 218 These modules enable custom tool definitions that run in-process via MCP

+4 -4

lib/client.mli

··· 36 36 {2 Message Flow} 37 37 38 38 1. Create a client with {!create} 2. Send messages with {!query} or 39 39 - {!send_message} 3. Receive responses with {!receive} or {!receive_all} 4. 39 39 + {!Advanced.send_message} 3. Receive responses with {!receive} or {!receive_all} 4. 40 40 Continue multi-turn conversations by sending more messages 5. Client 41 41 automatically cleans up when the switch exits 42 42 ··· 304 304 (** [receive_raw t] returns a lazy sequence of raw incoming messages. 305 305 306 306 This includes all message types before Response conversion: 307 307 - - {!Incoming.Message} - Regular messages 308 308 - - {!Incoming.Control_response} - Control responses (normally handled 307 307 + - {!Proto.Incoming.t.constructor-Message} - Regular messages 308 308 + - {!Proto.Incoming.t.constructor-Control_response} - Control responses (normally handled 309 309 internally) 310 310 - - {!Incoming.Control_request} - Control requests (normally handled 310 310 + - {!Proto.Incoming.t.constructor-Control_request} - Control requests (normally handled 311 311 internally) 312 312 313 313 Most users should use {!receive} or {!run} instead. *)

+1 -1

proto/structured_output.mli

··· 11 11 the specified JSON format, validated against the provided schema. 12 12 13 13 This is the protocol-level module. For the high-level API with logging and 14 14 - additional features, see {!Claudeio.Structured_output}. *) 14 14 + additional features, see {!Claude.Structured_output}. *) 15 15 16 16 (** {1 Output Format Configuration} *) 17 17