OpenAI Docs Guide

Query OpenAI official documentation directly via WebFetch.

When to Use

When the user asks about or the conversation involves:

You MUST WebFetch official documentation - never answer from memory!

Base URL: https://developers.openai.com/docs

Core Concepts:

Topic	URL
Text generation	/docs/guides/text
Code generation	/docs/guides/code-generation
Images & vision	/docs/guides/images-vision
Audio	/docs/guides/audio
Structured outputs	/docs/guides/structured-outputs
Function calling	/docs/guides/function-calling
Migrate to Responses API	/docs/guides/migrate-to-responses

Models & Pricing:

Agents:

Tools:

Topic	URL
Tools overview	/docs/guides/tools
Web search	/docs/guides/tools-web-search
File search	/docs/guides/tools-file-search
Code interpreter	/docs/guides/tools-code-interpreter
Image generation tool	/docs/guides/tools-image-generation
Computer use	/docs/guides/tools-computer-use
MCP connectors	/docs/guides/tools-connectors-mcp
Local shell	/docs/guides/tools-local-shell

Realtime API:

Topic	URL
Realtime overview	/docs/guides/realtime
WebRTC	/docs/guides/realtime-webrtc
WebSocket	/docs/guides/realtime-websocket
SIP	/docs/guides/realtime-sip
Transcription	/docs/guides/realtime-transcription

Run & Scale:

Topic	URL
Conversation state	/docs/guides/conversation-state
Streaming	/docs/guides/streaming-responses
Background tasks	/docs/guides/background
Prompt caching	/docs/guides/prompt-caching
Prompt engineering	/docs/guides/prompt-engineering
Reasoning	/docs/guides/reasoning
Webhooks	/docs/guides/webhooks

Fine-tuning & Optimization:

Topic	URL
Model optimization	/docs/guides/model-optimization
Supervised fine-tuning	/docs/guides/supervised-fine-tuning
Vision fine-tuning	/docs/guides/vision-fine-tuning
DPO	/docs/guides/direct-preference-optimization
Reinforcement fine-tuning	/docs/guides/reinforcement-fine-tuning

Specialized Models:

Topic	URL
Image generation	/docs/guides/image-generation
Video generation	/docs/guides/video-generation
Text-to-speech	/docs/guides/text-to-speech
Speech-to-text	/docs/guides/speech-to-text
Deep research	/docs/guides/deep-research
Embeddings	/docs/guides/embeddings
Moderation	/docs/guides/moderation

Production:

Topic	URL
Production best practices	/docs/guides/production-best-practices
Latency optimization	/docs/guides/latency-optimization
Cost optimization	/docs/guides/cost-optimization
Batch API	/docs/guides/batch
Safety best practices	/docs/guides/safety-best-practices

API Reference (for endpoint details):

Topic	URL
Responses API	/docs/api-reference/responses
Chat Completions	/docs/api-reference/chat
Images	/docs/api-reference/images
Audio	/docs/api-reference/audio
Embeddings	/docs/api-reference/embeddings
Fine-tuning	/docs/api-reference/fine-tuning
Files	/docs/api-reference/files
Models	/docs/api-reference/models

Prepend https://developers.openai.com to the path:

WebFetch("https://developers.openai.com/docs/guides/function-calling", "Extract the documentation content about...")

Extract relevant information from WebFetch results and answer the user directly.

If you can't find the right URL:

Try the mcp__openai-docs__search_openai_docs tool (search works, only fetch is broken)
Use the URL from search results with WebFetch
Fall back to WebFetch("https://developers.openai.com/docs/overview", "...") for the main index

Never answer OpenAI API questions from memory - always WebFetch first
The mcp__openai-docs__fetch_openai_doc tool is broken (returns 404) - do NOT use it
The mcp__openai-docs__search_openai_docs tool works fine for discovery
The mcp__openai-docs__get_openapi_spec tool works fine for API endpoint specs