OpenAI adds Model Context Protocol support to ChatGPT deep research
Custom connectors were limited to two read-only operations, search and fetch, rather than the full read-write access MCP allows — a restriction OpenAI lifted later that year.
- Models & capabilities
- Minor
OpenAI extended its support for the Model Context Protocol (MCP), an open standard for connecting AI models to external data sources and tools that Anthropic had published in November 2024, to ChatGPT’s deep research feature. Sam Altman had announced broader OpenAI support for MCP across its Agents SDK and Responses API in March 2025; the deep research integration let users and organisations point the tool at custom MCP servers rather than relying solely on OpenAI’s built-in connectors.
The initial implementation was narrow. Custom MCP servers used with deep research were required to expose just two read-only tools, search and fetch, following a compatibility schema OpenAI published alongside the feature — a deliberate restriction rather than the full read-write tool access MCP’s specification allows. That kept early third-party connectors limited to retrieving and citing external documents rather than taking actions through them.
The adoption was notable mainly as a concession: a protocol designed by a direct competitor became the industry’s de facto standard for tool-calling within months of release, and OpenAI’s uptake of it — alongside Google’s — did more than any single technical feature to establish MCP over OpenAI’s own, less widely adopted plugin systems from 2023. OpenAI later broadened MCP support considerably, adding full read-write “developer mode” access in September 2025.