Replicate

Summary

Replicate is a platform for running and sharing machine learning models through a simple API without managing infrastructure.

Release history

1. snapshot-20260820 seen Aug 20, 2026 · issue 002

   Replicate adds agent skills, MCP auto-discovery, prediction deadlines, fallback models, and source-filtered prediction listing.

   └──▷ TRY IT
   List only predictions you ran manually in the playground, separate from programmatic API runs.
   copy

   $ curl -s \
     -H "Authorization: Bearer $REPLICATE_API_TOKEN" \
     "https://api.replicate.com/v1/predictions?source=web"

   Add Replicate agent skills to a project so your coding assistant has expert guidance on model selection and prompting.
   copy

   $ npx skills add replicate/skills

   • ›Adds npx skills add replicate/skills to install Replicate agent skills — markdown instruction files giving coding assistants (Claude Code, OpenCode, OpenAI Codex) expert knowledge on model discovery, comparison, prompting, and API execution.
   • ›Adds allow_fallback_model parameter to Nano Banana Pro API calls to automatically fall back to ByteDance Seedream 5.0 lite when Google's API is rate-limited; the resolution field in output returns 'fallback' when triggered.
   • ›Adds /.well-known/mcp/server.json endpoint to Replicate's MCP server, enabling automatic discovery via the official MCP Registry (backed by Anthropic, GitHub, and Microsoft).
   • ›Exposes --tools flag on the MCP server metadata, letting clients choose between all (standard tools) or code (code mode) when installing.
   • ›Enables chat.mcp.gallery.enabled setting in VS Code to browse and install Replicate's MCP server directly from the Extensions view via @mcp search.
   +12 moreshow less

   • ›Adds source query parameter to the GET /v1/predictions list endpoint, accepting value web to filter predictions created through the Replicate web interface (results limited to the last 14 days).
   • ›Adds --tools=code flag to the local MCP server (npx -y replicate-mcp@alpha --tools=code) for experimental code mode, exposing two tools: SDK docs search and TypeScript code execution inside a Deno sandbox.
   • ›Adds Cancel-After request header on prediction creation to set a deadline duration (e.g. 1m) after which incomplete predictions are automatically cancelled.
   • ›Adds source field to prediction API responses indicating whether a prediction was created via web or api.
   • ›Launches deployment setup monitoring for enterprise customers: automatic email notifications on setup failure and configurable setup timeouts (default 10 minutes).
   • ›Adds approximate cost display to predictions and trainings on the dashboard and predictions pages.
   • ›Adds video support to the playground before/after slider for side-by-side comparison of video model outputs.
   • ›Adds pixelated image rendering with size controls (1x, 2x, fit) in the playground for pixel art models.
   • ›Adds search to the Deployments page.
   • ›Adds created and last updated dates on model pages.
   • ›Adds FAQs to every collection page.
   • ›Adds automatic llms.txt generation for Replicate documentation to aid language model discovery.