[DOCS] Add Ollama integration guide to Third Party Integrations / LLM Gateway sections

Documentation Type

Missing documentation (feature not documented)

Documentation Location

• https://code.claude.com/docs/en/third-party-integrations - https://code.claude.com/docs/en/llm-gateway

Section/Topic

• Third Party Integrations: The "Compare deployment options" table and general list of providers. - LLM Gateway Configuration: The section discussing compatible API formats and configuration examples.

Current Documentation

Currently, docs/en/third-party-integrations.md lists the following:

"Organizations can deploy Claude Code through Anthropic directly or through a cloud provider... Amazon Bedrock, Google Vertex AI, Microsoft Foundry."

And docs/en/llm-gateway.md mentions:

"For an LLM gateway to work with Claude Code, it must meet the following requirements... 1. Anthropic Messages: /v1/messages..." "LiteLLM configuration... Note: LiteLLM is a third-party proxy service."

What's Wrong or Missing?

Ollama (v0.14.0+) recently announced full compatibility with the Anthropic Messages API, explicitly enabling support for Claude Code with local open-source models (like qwen3-coder or gpt-oss).

The current documentation implies that only Cloud Providers (AWS, Google, Azure) or LiteLLM are supported. It is missing instructions for users who wish to use Claude Code with local inference servers like Ollama, despite this being a supported workflow via the standard ANTHROPIC_BASE_URL configuration.

Suggested Improvement

I suggest adding an "Ollama / Local Inference" section to LLM Gateway Configuration or Third Party Integrations.

Suggested Text:

Ollama Configuration

Claude Code can connect to local open-source models running via Ollama (v0.14.0 or later) using the Anthropic-compatible API.

1. Set Environment Variables
Configure Claude Code to point to your local Ollama instance:

export ANTHROPIC_BASE_URL=http://localhost:11434
export ANTHROPIC_AUTH_TOKEN=ollama

2. Run Claude Code
Launch Claude Code specifying an Ollama-supported model:

claude --model qwen3-coder

Note: It is recommended to use models with at least 32k context length for best performance with Claude Code.

Impact

High - Prevents users from using a feature

Additional Context

• Ollama Announcement Blog: https://ollama.com/blog/claude
• Ollama Documentation for Claude Code: https://docs.ollama.com/integrations/claude-code
• This integration allows users to run Claude Code's agentic loop entirely locally or against self-hosted infrastructure, which is a significant use case not currently covered in the docs.