Replicate

Replicate released a hosted remote MCP server that exposes model discovery and its HTTP APIs to clients including Claude Desktop, Claude Code, Cursor and VS Code, creating a specific agent-client distribution path beyond its official-model API catalog.

Replicate operates a production API for running, training and deploying public and private models, including more than 100 official models with stable APIs.

Replicate operates a production API for running, training and deploying public and private models, including more than 100 official models with stable APIs.

Replicate documents Cog packaging using cog.yaml for dependencies and predict.py for the prediction interface, generating Docker images and selecting CUDA and cuDNN versions for GPU frameworks before local testing and cloud publication.

Replicate provides model hosting, deployments, training, hardware abstraction and stable inference APIs across many model families.

Replicate provides model hosting, deployments, training, hardware abstraction and stable inference APIs across many model families.

Replicate and Cloudflare announced on November 17, 2025 that Replicate was joining Cloudflare and welcomed the team into Cloudflare. This corporate backing event is independent of existing inference-platform capabilities.

Replicate operates a mature developer infrastructure platform with maintained official models, deployments, training APIs and hardware abstraction.

Replicate operates a mature developer infrastructure platform with maintained official models, deployments, training APIs and hardware abstraction.

Replicate combines Cog model packaging, arbitrary GPU code execution, A100/H100/B200-class hardware, autoscaling from zero to thousands of GPUs, production monitoring, private endpoints, rolling deployments and GitHub-based CI/CD, spanning ML systems, cloud infrastructure, developer tooling and production AI operations.

Model serving, inference, training infrastructure and developer-platform skills transfer exceptionally broadly across AI infrastructure and model companies.

Model serving, inference, training infrastructure and developer-platform skills transfer exceptionally broadly across AI infrastructure and model companies.