Mistral Vibe: Built-in Web Search and Fetch Tools
Mistral VibeMistral Vibe 2.3.0 adds native web search and web fetch tools directly to the coding agent, allowing it to retrieve live web content and search results without requiring an external MCP server. The agent can now look up documentation, research dependencies, and fetch URLs as part of its standard toolkit during any coding session. This closes a long-standing gap between Mistral Vibe and browser-integrated AI tools, giving the terminal agent genuine awareness of the live web.
Sources & Mentions
3 external resources covering this update
Built-in Web Search and Fetch Tools
Mistral Vibe 2.3.0 brings native web search and web fetch tools directly into the coding agent, eliminating the need for an external MCP server to access the internet.
Web Search Tool
The agent can now execute inline search queries and reason over the results within its workflow. When you ask Vibe to research a library, look up error messages, or find recent API changes, it issues a search query, receives ranked results, and incorporates the findings into its response — all without leaving the terminal.
Web Fetch Tool
The web fetch tool retrieves full content from any URL: documentation pages, GitHub gists, README files, blog posts, and more. Vibe reads the page, converts it to a usable format, and uses the content to inform its coding decisions. This is especially useful when the agent needs to reference external documentation or pull in example code.
Why This Matters
Browser-integrated AI tools like Cursor and Windsurf have long had the advantage of built-in web awareness. With this release, Mistral Vibe closes that gap for terminal-based workflows. Developers no longer need to switch contexts or configure MCP servers just to give their agent access to live information.
Tool Call Streaming
The release also introduces incremental argument rendering for tool calls. Instead of waiting for a tool call to complete before displaying anything, Vibe now streams the arguments as they are generated, giving users real-time visibility into what the agent is doing.