// Agent-first setup
Set up Upwarden with your coding agent
Create one API key, paste one prompt, let your agent inspect your app, ask the right questions, plan the setup, and configure monitoring through the Upwarden API.
One guided prompt
You are setting up Upwarden for this application. First read: https://upwarden.eu/llms.txt API key: <UPWARDEN_API_KEY> Your job: 1. Inspect this codebase before making changes. 2. Identify the application type, production entrypoints, existing health checks, background jobs, APIs, queues, cron jobs, and deployment setup. 3. Do not create anything yet. 4. Ask me concise questions for anything that cannot be inferred safely. 5. Propose a monitoring and status-page setup plan with options. 6. Wait for my approval. 7. After approval, use the Upwarden REST API to create the agreed setup. 8. If needed, add or adjust health endpoints or heartbeat pings in this codebase. 9. Run the project's relevant tests/checks. 10. Return the status page URL, monitor list, heartbeat URLs, and any code changes made. Default to safe, minimal monitoring. Do not expose the API key in files, commits, logs, or screenshots.
Inspect first
The agent reads your repository and detects services, routes, jobs, health checks, and deployment clues before proposing anything.
Plan with you
It asks only for missing high-impact decisions, then shows monitoring options and waits for approval.
Execute safely
After approval it uses the REST API, keeps the key out of files, and verifies the created setup.
What your agent should discover
- Framework and runtime
- Production URL
- Health endpoint
- Public web app
- API endpoints
- Background jobs
- Queues
- Cron jobs
- WebSocket endpoints
- Externally monitorable database or cache dependencies
- Deployment platform
Questions your agent may ask
- Which production domain should Upwarden monitor?
- Should the status page be public?
- Which services should customers see as components?
- Do you want heartbeat monitoring for background jobs?
- Should I add a /health endpoint if none exists?
- Which check interval should I use within your plan limits?
// Runbook
A confidence workflow, not a blind recipe
The agent should fit Upwarden into the user's application and deployment model. It should not force every project into the same setup or create resources before the user has seen the plan.
What gets created after approval
- Status page
- HTTP monitors
- Optional WebSocket, TCP, ping, DNS, or domain monitors
- Heartbeat monitors
- Status-page components linked to monitors
- Optional health endpoint or heartbeat hooks in your codebase
Agent pseudocode
read("https://upwarden.eu/llms.txt")
apiKey = ask_user_or_read_secret("UPWARDEN_API_KEY")
assert apiKey starts with "upw_"
repo = inspect_codebase_read_only()
app = infer_application_shape(repo)
candidates = {
healthEndpoints: find_health_endpoints(repo),
publicServices: find_public_routes_or_apps(repo),
backgroundJobs: find_cron_jobs_queues_workers(repo),
websocketEndpoints: find_websocket_servers(repo),
deployTargets: find_deployment_config(repo)
}
questions = build_questions_for_missing_high_impact_info(candidates)
if questions not empty:
ask_user(questions)
plan = propose_upwarden_setup(candidates, user_answers)
show_plan(plan)
wait_for_user_approval()
validate_api_key()
existing = list_existing_upwarden_resources()
resources = create_or_reuse_resources(plan, existing)
if plan.requires_code_changes:
implement_health_or_heartbeat_changes()
run_project_tests()
verify_created_resources(resources)
return_summary(resources)