agent-framework-hosting-mcp

# agent-framework-hosting-mcp MCP conversion helpers for app-owned Agent Framework hosting. The package deliberately does not choose a web framework or wrap the MCP SDK server lifecycle. It provides two conversion functions and small adapters: - `mcp_to_run(...)` converts native MCP tool arguments into Agent Framework run arguments. - `mcp_from_run(...)` converts an `AgentResponse` or `Message` into native MCP `ContentBlock` values. - `AgentMCPTool(...)` generates the native `Tool` definition from an agent and keeps listing, parsing, execution, result conversion, and optional `AgentState` session persistence aligned. - `WorkflowMCPTool(...)` generates the native `Tool` definition from a workflow's start-executor input type and converts completed workflow outputs. Application code keeps ownership of the MCP SDK's `Server`, handler registration, request context, transport, session-key policy, authentication, authorization, and deployment. For direct conversion, the argument name is part of the app-owned MCP tool contract. Define it once and use the same value in both the native tool schema and `mcp_to_run(...)`: ```python agent_input_argument = "task" chat_option_arguments = { "reasoning_effort": { "type": "string", "enum": ["low", "medium", "high"], }, } tool = Tool( name="run_agent", inputSchema={ "type": "object", "properties": { agent_input_argument: {"type": "string"}, **chat_option_arguments, }, "required": [agent_input_argument], }, ) run = mcp_to_run( arguments, argument_name=agent_input_argument, chat_option_arguments=chat_option_arguments, ) ``` Only names listed in `chat_option_arguments` are copied to `run["options"]`; other MCP arguments remain available in the message's raw representation but are not forwarded to the model client. The native MCP schema remains responsible for validating exposed option types and ranges. For an agent …