Google Brings Its Developer Documentation Into the Age of AI Agents

"Language models trained on fixed documentation will quickly fall out of sync with fast-changing platforms. When AI assistants confidently generate code against deprecated APIs or missing features, the resulting bugs can be subtle and costly to debug. The Developer Knowledge API acts as a programmatic source of truth for Google's public documentation."

"MCP is an open standard that many in the industry are adopting. It lets AI assistants safely access external data sources in real-time instead of just using their built-in training knowledge. The practical upshot is that an AI assistant can now look up the authoritative answer to questions rather than hallucinating a plausible-sounding but incorrect one."

Google introduced the Developer Knowledge API public preview with a Model Context Protocol server to address a critical challenge in AI development: language models generating code based on outdated or incorrect documentation. The API provides two core functions—SearchDocumentChunks for finding relevant documentation snippets and GetDocument/BatchGetDocuments for retrieving full content. The accompanying MCP server, an open industry standard, allows AI assistants to access current Google documentation in real-time rather than relying solely on training data. This prevents costly bugs from deprecated APIs and missing features. The search_document tool enables natural language queries, while get_document retrieves specific page content, ensuring AI assistants provide authoritative answers instead of generating plausible but incorrect information.

Read at InfoQ