Refactor Agents

docs/changelog.mdx

@@ -25,7 +25,7 @@ description: "Product updates and release notes for HUD SDK and Platform."
 - **`hud sync env`** — sync local environment configs with collision detection (replaces `hud link`).
 - **`hud eval` accepts Python files** — run evaluations directly from `.py` files and directories containing `Task` objects.
 - **Chat class** — manage multi-turn agent conversations from a single SDK abstraction.
-- **GPT-5 support** — `ResponseAgent` defaults to `gpt-5`, with ToolSearch tool support.
+- **GPT-5 support** — auto-response classification defaults to `gpt-5`, with ToolSearch tool support.
 - **Citations** — citation support for Claude, Gemini, and OpenAI responses in chat and agent traces.
 
 ### Platform

docs/reference/agents.mdx

@@ -42,7 +42,7 @@ Abstract base class for all MCP-enabled agents. Handles the agent loop, MCP clie
 |-----------|------|-------------|---------|
 | `mcp_client` | `AgentMCPClient` | MCP client for server connections | `None` |
 | `auto_trace` | `bool` | Enable automatic tracing spans | `True` |
-| `auto_respond` | `bool` | Use ResponseAgent to decide when to stop/continue | `False` |
+| `auto_respond` | `bool` | Use response automation to decide when to stop/continue | `False` |
 | `verbose` | `bool` | Verbose console logs for development | `False` |
 
 **Base Config** (shared by all agents):
@@ -63,9 +63,6 @@ async def run(ctx: EvalContext, max_steps: int = 10) -> Trace:
 
 async def call_tools(tool_call: MCPToolCall | list[MCPToolCall]) -> list[MCPToolResult]:
     """Execute tool calls through MCP client."""
-
-def get_available_tools() -> list[types.Tool]:
-    """Get filtered list of available tools."""
 ```
 
 ## Pre-built Agents
@@ -251,7 +248,7 @@ result = await agent.run(task, max_steps=20)
 
 ### Auto-Respond Mode
 
-When `auto_respond=True`, the agent uses a ResponseAgent to decide whether to continue or stop after each model response:
+When `auto_respond=True`, the agent uses response automation to decide whether to continue or stop after each model response:
 
 ```python
 agent = ClaudeAgent.create(

docs/reference/cli/eval.mdx

@@ -79,7 +79,7 @@ hud eval [SOURCE] [AGENT] [OPTIONS]
 </ParamField>
 
 <ParamField path="--auto-respond" type="boolean">
-  Use ResponseAgent to decide when to stop/continue. Default: True for `--full`.
+  Use response automation to decide when to stop/continue. Default: True for `--full`.
 </ParamField>
 
 ### Taskset Association

docs/reference/types.mdx

@@ -111,12 +111,12 @@ print(result.reward, result.done)
 | `trace` | `list[TraceStep]` | Execution trace steps |
 | `messages` | `list[Any]` | Final conversation state |
 
-## InferenceResult
+## AgentResponse
 
 Returned by agent `get_response()` methods. Represents the result of a single LLM inference call.
 
 ```python
-from hud.types import InferenceResult
+from hud.types import AgentResponse
 ```
 
 | Field | Type | Description |
@@ -129,8 +129,6 @@ from hud.types import InferenceResult
 | `info` | `dict[str, Any]` | Provider-specific metadata |
 | `isError` | `bool` | Error flag |
 
-> **Note:** `AgentResponse` is available as a backwards-compatible alias for `InferenceResult`.
-
 ## AgentType
 
 Enum of supported agent types.

hud/agents/__init__.py

@@ -1,95 +1,15 @@
 from __future__ import annotations
 
-import sys
-from types import ModuleType
-from typing import Any
-
-from .base import CategorizedTools, MCPAgent
+from .base import MCPAgent
 from .claude import ClaudeAgent
+from .gateway import create_agent
 from .openai import OpenAIAgent
 from .openai_compatible import OpenAIChatAgent
 
 __all__ = [
-    "CategorizedTools",
     "ClaudeAgent",
     "MCPAgent",
     "OpenAIAgent",
     "OpenAIChatAgent",
     "create_agent",
 ]
-
-
-def _install_openai_chat_compat_module() -> None:
-    module_name = f"{__name__}.openai_chat"
-    if module_name in sys.modules:
-        return
-
-    module: Any = ModuleType(module_name, "Compatibility module for OpenAIChatAgent.")
-    module.OpenAIChatAgent = OpenAIChatAgent
-    module.__all__ = ["OpenAIChatAgent"]
-    sys.modules[module_name] = module
-
-
-_install_openai_chat_compat_module()
-
-
-def create_agent(model: str, **kwargs: Any) -> MCPAgent:
-    """Create an agent for a gateway model.
-
-    This routes ALL requests through the HUD gateway. For direct API access
-    (using your own API keys), use the agent classes directly.
-
-    Args:
-        model: Model name (e.g., "gpt-5.4", "claude-sonnet-4-6").
-        **kwargs: Additional params passed to agent.create().
-
-    Returns:
-        Configured MCPAgent instance with gateway routing.
-
-    Example:
-        ```python
-        # Gateway routing (recommended)
-        agent = create_agent("gpt-5.4")
-        agent = create_agent("claude-sonnet-4-6", temperature=0.7)
-
-        # Direct API access (use agent classes)
-        from hud.agents.claude import ClaudeAgent
-
-        agent = ClaudeAgent.create(model="claude-sonnet-4-6")
-        ```
-    """
-    from hud.agents.gateway import build_gateway_client
-    from hud.agents.resolver import resolve_cls
-
-    # Resolve class and gateway info
-    agent_cls, gateway_info = resolve_cls(model)
-
-    # Get model name from gateway info or use input
-    model_id = model
-    if gateway_info:
-        model_id = gateway_info.get("model_name") or model
-
-    # Determine provider: from gateway info, or infer from agent class
-    if gateway_info:
-        provider = gateway_info["provider"]["name"]
-    else:
-        provider = "openai"
-        if agent_cls.__name__ == "ClaudeAgent":
-            provider = "anthropic"
-        elif agent_cls.__name__ == "GeminiAgent":
-            provider = "gemini"
-
-    client = build_gateway_client(provider)
-
-    # Set up kwargs
-    kwargs.setdefault("model", model_id)
-
-    # Use correct client key based on agent type
-    if agent_cls == OpenAIChatAgent:
-        kwargs.setdefault("openai_client", client)
-    else:
-        # Claude and other agents use model_client and validate_api_key
-        kwargs.setdefault("model_client", client)
-        kwargs.setdefault("validate_api_key", False)
-
-    return agent_cls.create(**kwargs)