Module operator 

Operator MCP server injection.

Every non-internal agent in Construct gets the Operator orchestration MCP server wired in automatically. This module defines:

• The canonical McpServerConfig for the Operator stdio server.
• The system-prompt text that teaches the lead agent how to use operator tools (create_agent, wait_for_agent, etc.).
• inject_operator() – called during agent/config construction to splice both the server config and the prompt into whatever Config is being assembled.

Multi-Provider Support

The operator prompt is split into two layers:

1. Core layer (core::OPERATOR_CORE_PROMPT) — universal orchestration philosophy that works with any LLM (planning, governance, rules).
2. Tool layer (providers) — provider-specific tool-calling examples adapted for each LLM family (Claude MCP, OpenAI JSON, etc.).

Provider detection is automatic via providers::Provider::detect using the model name string.

Modules

core

Universal operator prompt — provider-agnostic orchestration philosophy.

providers

Provider detection and tool-layer dispatch.

Constants

DEFAULT_OPERATOR_MCP_PATH_SUFFIX

Default path to the Operator MCP runner script (relative to $HOME).

OPERATOR_PROMPT

Backward-compatible constant — use build_operator_prompt instead.

OPERATOR_SERVER_NAME

Name used as the MCP server prefix (tools appear as construct-operator__<tool>).

Functions

append_operator_channel_prompt

Append the compact operator prompt for channel agents.

append_operator_prompt

Append the full Operator prompt to system_prompt.

build_operator_prompt

Build the complete operator prompt for a given model.

inject_operator

Inject the Operator MCP server into config.

operator_mcp_server_config

Build the McpServerConfig for the Operator stdio server.

resolve_operator_mcp_path

Resolve the absolute path to run_operator_mcp.py.