Replicate now publishes agent skills, a collection of markdown instruction files that give coding assistants expert knowledge about working with AI models on Replicate. Skills cover model discovery, comparison, and execution via the API, along with detailed prompting techniques for image generation and video generation models.

Nano Banana Pro can now fall back to a backup model when Google’s API is at capacity, instead of failing with a rate limit error. Set `allow_fallback_model` to `true` when calling the API to use Seedream 5.0 lite as a fallback.

Replicate’s MCP server can now be discovered automatically through the official MCP Registry. Added a `/.well-known/mcp/server.json` endpoint to publish metadata about the MCP server, following the server.json specification from the Model Context Protocol.

You can now filter the list predictions API endpoint to show only predictions created through the web interface using the `source` query parameter with a value of `web`. This is useful to see predictions created using the playground or other parts of the Replicate website.

Improved the reliability of google/nano-banana and google/nano-banana-pro, improved accessibility when using the search bar across Replicate, and added automatic llms.txt generation for documentation.

Added created and last updated dates on model pages, improved pricing display on model pages, and made FLUX.2 [pro] the default model in the playground.

Added approximate cost display to predictions and trainings, launched an updated homepage hero section, and published a blog post on how to prompt Nano Banana Pro.

Replicate’s local MCP server now supports an experimental 'code mode' that allows language models to write and execute TypeScript code directly in a sandboxed environment. This mode is designed for more efficient complex workflows.

Added video support to the before/after slider, added pixelated image rendering size controls in the playground, and improved support for aspect ratios in prediction outputs.

Customers on an enterprise plan now have more control over their deployments with automatic notifications when deployments fail setup and customizable setup timeouts.