Start
Prompts
Copyable prompts for handing a Userland app build to a coding agent, or to ChatGPT, Claude or Grok Bot with Userland connected.
For agents: Use the full prompt as an operating contract. Report files created, commands run, app identity, activation status, and rollback notes.
App Build Prompt
Read https://docs.userland.fun/llms.txt and build a native Cloudflare Worker app for this idea: <APP_IDEA>.
Start with Userland context discovery. Use my selected Account, existing authority and actual capabilities. Retrieve existing app state before choosing whether to create or update. Author standard Worker source, wrangler.jsonc, native bindings and app-owned authentication; do not use a Userland runtime SDK or custom manifest.
Apply source under an exact app revision and stable idempotency key. Poll the returned operation, inspect build/runtime failures, repair and retry safely. Use the hosted browser when available to verify the requested essential flows and saved data. A successful build, screenshot or HTTP200 is not enough.
Publish publicly if this request authorizes that audience; otherwise keep a protected preview while continuing independent work. Ask only for a truly missing decision, external credential or permission. Keep secret values out of source and messages; generated app secrets should not require me to copy a value.
Report the app address, Account/app identity, deployment state, checks performed, unresolved dependencies and recovery options. Preserve source and operation identities for the next conversation.
Update Prompt
Read https://docs.userland.fun/llms.txt. Update my native Userland app <APP_ID>: <CHANGE>.
Discover current Account authority and capabilities, then retrieve the selected environment's state and complete source. Use exact revisions and per-file hashes for an atomic patch; do not reconstruct the project from chat memory. Apply, inspect and repair the change, and verify the affected essential flows.
Continue within my existing authorization without a redundant Publish prompt. Do not silently broaden the audience, spending or credential scope. Report the active deployment, meaningful checks and any real unfinished dependency.
Chat Build Prompt
Use the Userland connector to build a native app for this idea: <APP_IDEA>.
Read the complete native guide, starting at userland://guide/native/index, or https://docs.userland.fun/llms.txt. Call context_get, then use apps_get, apps_apply, apps_run and operations_get as needed. Discover capabilities before assuming hosted builds, browsers or bindings are configured.
Use normal Cloudflare Worker code and app-owned authentication. Apply, inspect, test and repair through the connector without sending me to a terminal. Keep credentials out of source and chat. Generate app secrets through typed requirements; use a protected Userland browser action only for a genuinely missing external credential.
Publish publicly when my request authorizes it; otherwise use a protected preview and continue independent work. Report the working link, checks performed, Account/app identity, unresolved dependencies and recovery options. Store enough state for a fresh conversation to continue.
Chat Update Prompt
Use the Userland connector to update my native app <APP_ID>: <CHANGE>.
Read the native guide and current context. Retrieve the complete saved source, exact revision and selected environment, then make an atomic patch. Apply it, inspect the operation and verify the changed behavior with the advertised hosted tools. Repair a failed build or app check before claiming success.
Use existing authorization without asking for another Userland Publish approval. Never request secret values in chat. Report the active deployment, checks performed and any actual blocker; a new conversation should be able to recover the same app from Userland.
Userland docsYours, live.