#751 call_llm implements only Ollama; Anthropic/OpenAI raise NotImplementedError (blocks Claude support)

engine/tools/llm/core/providers.py:87-112 — _call_openai/_call_anthropic/_call_grok/_call_gemini/_call_qwen all raise NotImplementedError, and their stub signatures take only prompt (no messages, no tools). Highway Agents promises 'Ollama or Claude', so Claude needs a real Anthropic provider: Messages API with multi-turn messages + tools/tool_choice (function/tool-use), mapped to the existing return shape (response, finish_reason, usage, tool_calls). Until then Phase 0 / tools.agent.run_goal is Ollama-only and must use a tool-calling-capable Ollama model. Relates #749.