# Flarely

> Hosted Playwright checks, agent first. A coding agent writes a check over MCP, Flarely runs it in a container on Cloudflare, keeps the trace, and returns a failure context (failing test, step, locator, expected/received, artifacts) tight enough to fix the check from. Then the check runs on a schedule and on deploy, with incidents and alerts.

## Start
- Sign up: https://app.flarely.app/sign-up
- MCP server: https://mcp.flarely.app/mcp with header "Authorization: Bearer fl_api_..."
- API base: https://api.flarely.app/v1

## Docs
- Getting started: https://flarely.app/docs
- Writing checks: https://flarely.app/docs/checks
- MCP for coding agents: https://flarely.app/docs/mcp
- CLI: https://flarely.app/docs/cli
- GitHub Action: https://flarely.app/docs/github-action
- Deploy hooks: https://flarely.app/docs/deploy-hooks
- Signed traffic and evidence bundles: https://flarely.app/docs/signed-traffic
- Bot info for site owners and WAF operators: https://flarely.app/bot
- Schedules and alerts: https://flarely.app/docs/schedules-and-alerts
- API: https://flarely.app/docs/api
- Pricing: https://flarely.app/pricing

## MCP tools
list_projects, list_checks, get_check, create_check, update_check, run_check (waits by default), get_run, get_failure_context, get_check_history, configure_check, list_incidents.

## Conventions for checks
Standard @playwright/test specs. Read the target from process.env.BASE_URL. One flow per check. Prefer getByRole and getByLabel. Wrap phases in test.step so the failure context can name where it broke. Assert visible outcomes.