Build with frameworks on Cloudflare

Use the framework as the primary development interface and Cloudflare Workers as the deployment target. Serve static sites and SPA output with Workers Static Assets. Do not select Pages templates, Pages deployment commands, or Pages-specific presets for new applications. Preserve existing Pages deployments during unrelated maintenance.

Route the project

1. Inspect the repository before changing it. Identify the framework, installed version, package manager, rendering mode, existing adapter, and Cloudflare configuration: cloudflare.config.ts, or wrangler.jsonc, wrangler.json, or wrangler.toml.
2. Preserve the user's framework choice. Determine whether the task starts a new project, adapts an existing project, or changes only application code.
3. Read only the matching row below, its Workers framework guide, and the framework's own agent guidance.
4. Set up the project for its state:
   • New project: follow the framework's Workers guide for project creation. The guides currently create the project with C3 and configure it for Wrangler; this is an exception to the cloudflare skill's default of cf for new projects, because the cf docs do not yet cover creating framework projects. After creation, the project has Wrangler configuration, so the next bullet applies.
   • Existing project with Wrangler configuration: keep Wrangler and load the wrangler skill. Follow the framework guide for adapter and configuration changes. Do not run cf init ., cf dev, cf build, or cf deploy in it.
   • Existing project with no Cloudflare configuration: use cf, which detects supported frameworks and configures them; see Start from an existing project and the cf-cli skill. Detection does not mean the project builds. If the build fails or the framework is not supported, follow the framework guide's manual setup.
5. Change only the adapter, runtime, build, and deployment configuration the integration requires. Inspect generated package scripts, run the framework's checks and build, then verify the Worker locally before deployment.