What's new in Checkly

Checkly CLI 10: see what a deploy will change before it happens — permalink

Before checkly deploy applies anything, it now shows you exactly what will change. If someone edited a check in the web app, the CLI can copy those edits back into your code, so you don't overwrite them by accident. Want to be strict about it? Add --fail-on-drift and the deploy stops if it would overwrite web app edits.

CLI 10 also brings a few quality-of-life updates. checkly login now works over SSH and for AI agents. Cancelling a test session also skips queued checks of every type. Checks that are already running finish. With CLI 10.1.0 and GitHub Action v1.2.0, the Re-run button on a Checkly GitHub Check re-runs your CI job.

Before you upgrade: CLI 10 has breaking changes. Deploys no longer start check runs unless you add --schedule-on-deploy. AgenticCheck and other deprecated APIs are removed. The release notes have the full list. Your first deploy after upgrading may show many updates. That's expected.

Upgrade with npm install -D checkly@latest.

Save your filters as views — permalink

Save any set of filters as a view. Views show up as tabs above Monitors and Test sessions, and you can keep them private or share them with your team. They replace the traffic-light tiles on the home dashboard, so the counts you see always follow your filters.

Finding the right checks is easier too. Monitors, Heartbeats and Test sessions get a new filter builder with and/or conditions. Reporting gets the same filters and views, plus a Chart view that compares up to 12 checks side by side.

AI Assistant on all paid plans and trials — permalink

The AI Assistant is now available on every paid plan and during trials. Describe what you want to monitor, and it tests the check before it creates it, so you go from a prompt to a working check.

Each new conversation uses one AI invocation. Trials now include 20 invocations, up from 10.

Automatic check repair (Beta) got smarter too. You can now start a repair yourself, Rocky won't make a repair that would weaken a check, and every repair shows up in the check's timeline.

See your checks right inside your AI client — permalink

The Checkly MCP server now shows interactive views in AI clients that support them, like Claude Desktop. Ask about your checks and get a status card, a check list or a response-time chart instead of plain text. Available for all accounts.

Reminder: Agentic Checks stop on December 1 — permalink

Agentic Checks stop running on December 1, and CLI 10 no longer supports them. If you still use them, switch to a Browser or Multistep check with automatic repair, or use the MCP server. The migration guide walks you through both.

Create and manage checks from your AI agent — permalink

The Checkly MCP server can now create, update, and delete checks, monitors, and check groups. Tell your agent what to monitor, and it sets up the check.

Here’s an example: Create a URL monitor for checklyhq.com that runs every minute from Europe and the US. Alert me if it goes down. If you dreamt about feeding your agent CSV files or OpenAPI specs to start monitoring all the things, now’s the time to send it to work!

Playwright Check Suites aren't supported yet because they need access to local files.

Automatic check repair for Browser and Multistep checks (Beta) — permalink

As mentioned above, setting up new checks is only one prompt away, but what do you do when checks fail? How can you evaluate and, in the ideal case, fix them?

Rocky AI can now repair browser and multistep checks for you if the root cause analysis finds a repairable mistake like an outdated selector. In that case, Rocky proposes a fix, tests it, and updates the check if it still matches the initial intent. It's off by default.

Track contract usage and credit budgets — permalink

Contract customers can now see consumed and remaining credits, usage trends, and projected consumption, broken down by account and check type, including AI invocations. Spot a projected overage early, without asking us for a report. The same data is available through the API, CLI, and MCP server.

Status pages, rebuilt around components — permalink

We rebuilt our status pages (say hello to StatusPageV3) and introduced components. The new component-based status pages make it easy for you to group your services and properties in the way your users will understand. Check them out!

And if you're on Team and Enterprise plans, automation rules can open and resolve incidents for you.

Private Locations now report their own errors — permalink

Debug private location errors without leaving Checkly. Private Locations get an Errors tab with an agent-errors overview, error counts over time, stack traces, breadcrumbs, and the agent ID and version behind every error. Node.js-side error reporting requires agent v8.7.0.

See, inspect, and clean up projects from the app — permalink

If your Checkly account holds monitors defined in different codebases, Monitoring as Code projects are now visible in the UI. Inspect which resources a project contains, review its deployment context (including failed Checkly deployments), navigate managed resources, and spot drift between code and account. And as a bonus: you can now preserve resources with their history if you remove them from your codebase.

Resolve invocations now follow your billing period — permalink

Rocky AI root cause analyses are counted as Resolve invocations against your billing period instead of calendar-month slices — yearly customers can use their allowance across the whole year. Need more? Additional invocation packs are now available to purchase.

Maintenance windows are now timezone-aware — permalink

Recurring maintenance windows keep their configured local time across daylight saving changes. No more manually shifting schedules twice a year.

Run Checkly from GitHub Actions with native CI feedback — permalink

The new checkly-action runs checkly test or checkly trigger straight from your workflow, with detached test sessions and results reported through GitHub Checks. Gate pull requests on Checkly results without burning Actions minutes waiting for runs to finish.

- uses: checkly/checkly-action@v1
  with:
    command: test

✓ 12 checks passed · reported via GitHub Checks

gRPC, SSL, and Traceroute monitors are generally available — permalink

Three new monitor types, now available to everyone: check gRPC endpoints, catch expiring certificates, and trace network paths. Create and manage them from the web app, API, CLI, or your Monitoring as Code setup, with detailed assertions and diagnostics.

Web Vitals in Playwright Check Suite results — permalink

Playwright Check Suites now collect six Web Vitals for every page your tests visit — LCP, CLS, INP, FCP, TTFB, and TBT, scored against the web.dev thresholds. No changes to your tests needed: metrics show up per page visit in the same result view you already use for functional failures, and each LCP and INP score expands into a breakdown of where the time went.

Status Pages now publish RSS, Atom, and Slack feeds — permalink

Every status page now exposes RSS and Atom feeds, and incident updates can flow straight into a Slack channel. Subscribers get incidents where they already work instead of refreshing a status page.

The Checkly MCP Server is here — permalink

The first version of the Checkly MCP Server is live. Point your AI client at the server URL and manage your monitoring from any MCP-capable agent — setup instructions per client are in the docs.

https://api.checklyhq.com/mcp

Download check artifacts straight from the terminal — permalink

No more digging through nested check-result JSON. checkly assets list and checkly assets download pull the artifacts of any check run or test session — logs, Playwright traces, videos, screenshots, reports — into your local folder.

Tag Manager: manage tags in one place — permalink

Rename tags everywhere they are used, delete unused or misspelled ones, and add tag descriptions that give Rocky AI extra context for root cause analyses. Available in the UI under Configuration.

The Checkly Slack app — permalink

Install the Checkly Slack app to receive alerts and rerun checks or AI analyses directly from Slack. It replaces webhook-based Slack alerts and comes with a new CLI construct.

More context for automatic root cause analysis — permalink

Rocky AI's root cause analysis now takes in two extra artifacts to determine why a check failed. It pulls in the last passing result for a failing check and diffs it against the failure to spot obvious regressions, and if you have Checkly Traces enabled, it pulls in relevant OpenTelemetry traces to correlate downstream errors with the failing check result.

Checkly CLI 8: ESM-only, cancellable runs, recording by default — permalink

CLI 8 ships three major changes: it is ESM-only (Node.js ^20.19.0 or >=22.12.0), CTRL+C now cancels test, pw-test, and trigger runs instead of burning quota in the background, and runs record by default unless you pass --no-record. Use -d for detached mode when you want a run to keep going.

npm i -D checkly@latest

Playwright Check Suites: faster runs, fewer false negatives — permalink

A batch of reliability work: cancellable queued and in-progress runs, custom runtime engines (Bun 1.7, Node.js 22/24/26), automatic retries for setup steps hit by transient network issues, browser caching in Private Locations, and full Playwright globalTimeout support.

Private Location metrics in the UI and Prometheus — permalink

See queue duration, agent count, and scheduled check runs for your Private Locations right in the UI — and spot over- or underprovisioned agents at a glance. The same metrics are available via Prometheus, so you can autoscale agents on Kubernetes with KEDA.

Alert channels now support environment secrets — permalink

Use {{MY_SECRET_VAR}} syntax in alert channel configs, so API keys and other sensitive values no longer live in plain text in the UI or API responses. Convert existing values with the new lock icon.

Cancel runs whenever you need to — permalink

Playwright Check Suite runs are now cancellable. You can stop queued or in-progress runs from the Playwright Check Suite Runs or Test Session Run pages, or press CTRL+C on Checkly CLI 8.0.0 and later. This works for runs started from Schedule now and for test sessions triggered via npx checkly test, npx checkly pw-test, and npx checkly trigger; cancelled runs show a Cancelled status in your run history.

Choose your Node/Bun runtime engine — permalink

Playwright Check Suites now support multiple Node.js and Bun engines, configurable from the UI or in code. Starting with Checkly CLI 8.1.0, the CLI reads your project's version files (.nvmrc, .node-version, .bun-version) and selects a compatible runtime automatically, with Node.js 22 (default), 24, and 26 and Bun 1.3 supported. You can override detection by setting the engine property on your PlaywrightCheck, for example engine: Engine.node('24').

Browser caching across runs in Private Locations — permalink

Check Suites now cache Playwright browsers across runs on the same Private Location agent, so the installation step is skipped entirely when a cached version exists. That means faster startup, no repeated downloads, and lower bandwidth usage, with zero configuration required. Caching is scoped per agent and available since Private Location agent version 8.2.0.

ESM-only and updated Node.js requirements — permalink

The Checkly CLI is now distributed as ESM only instead of CommonJS, and CLI 8 requires Node.js ^20.19.0 || >=22.12.0. Make sure your local development, CI, and deployment environments run a compatible Node.js version before upgrading; your own projects can still use CommonJS. The CLI also switches its TypeScript loader from ts-node to jiti, which is now bundled as a direct dependency.

npm install checkly@latest

CTRL+C now cancels running Playwright Check Suites — permalink

Pressing CTRL+C during test, pw-test, or trigger now cancels in-flight test sessions instead of leaving them running in the background. If you want the previous keep-running-in-the-cloud behavior, run in detached mode with --detach or -d.

--record now defaults to true — permalink

The test and trigger commands now record results by default. To disable recording, pass --no-record; if your scripts relied on recording being off by default, update them before upgrading.

Use the OpenAPI spec instead — permalink

The old API spec at api.checklyhq.com/swagger.json is deprecated and frozen. Use the newer OpenAPI spec at api.checklyhq.com/openapi.json instead. The old file remains available for compatibility, but now carries a deprecation notice in its description and response headers.

Every failing assertion gets its own error group — permalink

Failing assertions on API, URL, DNS, ICMP, and TCP checks now produce individual error messages in Error Groups, instead of being consolidated under a generic "1 of 3 assertions failed". This means Rocky AI root cause analysis triggers when a genuinely new assertion failure surfaces, rather than staying silent because the failures looked alike. Live for all users on all plans.

Get started with the Checkly CLI using your agent — permalink

The new checkly init command sets up Checkly Skills, provides a prompt tailored to your project, and one-shots your monitoring suite with Claude Code, Codex, Cursor, or any other agent of your choosing.

# From ~/path/to/my-unmonitored-project/
npx checkly init

Runtime 2026-04 is here — permalink

The new runtime is available for Browser and Multistep checks, with upgraded Node.js and Playwright versions and updated built-in dependencies. Upgrade from your account settings; Private Location support starts with agent v7.

Playwright Check Suites: Bun support and smarter dependency caching — permalink

Run suites on Bun from CLI 7.11.0 onwards. Custom package manager versions are respected down to the patch version, dependency caches upload only on successful runs and invalidate automatically when your lockfile changes, and all four package managers show detailed install logs.

Playwright 1.58.2 and Node.js 24.13.1 — permalink

Runtime 2026.04 brings Playwright v1.58.2 and Node.js v24.13.1 to Browser and MultiStep Checks. Head to Account settings > Runtimes to select it or make it your default; the runtime documentation lists the full spec with all 67 included packages.

Seven new dependencies — permalink

We added axios-cookiejar-support, csv-parse, imapflow, mysql2, pdf-parse, tough-cookie, and winston to the runtime, covering persistent cookie handling, CSV parsing, IMAP inbox monitoring, a modern MySQL client, PDF text extraction, and structured logging. We also removed jsdom; for DOM interactions, use Playwright's built-in Page APIs instead.

ICMP monitors join Checkly's stack — permalink

Checkly now supports ICMP monitors on all plans, letting you continuously ping hosts to measure reachability, latency, and packet loss across regions. This helps teams detect connectivity issues early and understand network performance before it impacts their applications.

Faster incident debugging with Rocky AI — permalink

Rocky AI, Checkly's AI agent, is now generally available on all plans. Its automated Root Cause Analysis analyzes failing checks and tests, investigates artifacts like traces and PCAP files, and summarizes the likely root cause so teams can debug issues faster.

Playwright Reporter 1.8: a major upgrade — permalink

The Checkly Playwright Reporter got a major overhaul. It now captures richer debugging data from Playwright runs, including traces, logs, network activity, and performance metrics, making it easier to debug CI failures without digging through artifacts.

Agent-friendly CLI — permalink

The Checkly CLI is now built for AI agents. Across v7.3.0 to v7.7.0 we shipped a skills-based self-discovery system via npx checkly skills, read-only commands to list and inspect checks, full incident lifecycle management from the terminal, and a confirmation protocol for safe agent-driven write operations. Recent additions include analytics stats from the terminal, a built-in checkly skills install command, and checkly account plan to let agents understand your plan's features.

Copy a prompt, get a working config — permalink

We added a Copy prompt button to the Playwright Check Suites onboarding. Paste the prompt into your AI coding tool (Claude Code, Cursor, Copilot) in the context of your repo: it reads your Playwright setup (projects, test files, base URLs, dependencies), generates check suite definitions grouped by project and tags, iterates until the config looks right, and deploys once you are ready. You get a working checkly.config.ts without writing one from scratch.

Also available for Test Sessions and the Playwright Reporter — permalink

The same copy-paste flow works for Test Sessions and Playwright Reporter setup, with a different prompt for each.

A production-ready reporter for Playwright — permalink

The Checkly Playwright Reporter runs alongside your Playwright tests, captures execution data (logs, traces, network activity, console output, performance metrics), and uploads it to Checkly so your team can debug runs without digging through CI artifacts or trace files. It works locally and in CI, and now automatically scrubs sensitive values from reports and trace files, with full control via scrubbing options like envVars, autoDetect, and replacement.

Getting started takes about a minute — permalink

Install the reporter, export CHECKLY_API_KEY and CHECKLY_ACCOUNT_ID, add createChecklyReporter() to the reporter list in your playwright.config.ts, and run npx playwright test. Then open the run in Checkly to explore traces, logs, network requests, console output, and performance data.

npm install --save-dev @checkly/playwright-reporter

ICMP monitoring is here — permalink

ICMP monitors are now available on all plans. They continuously ping your hosts with ICMP Echo Requests to measure reachability, latency, and packet loss across regions, so you can verify that hosts not exposed to the public internet stay reachable, catch network-level outages early, and compare latency geographically. Set your ping count and packet loss thresholds, and optionally add assertions on latency or packet loss.

Fully integrated with Monitoring as Code — permalink

ICMP monitors work in Checkly's Monitoring as Code workflows via the IcmpMonitor construct, available in the CLI from v7.1.0. Terraform support ships in provider v1.18.0 with the checkly_icmp_monitor resource; Pulumi (v2.8.0) is not yet supported.

Smarter alert notifications with Rocky AI — permalink

Rocky AI, Checkly's AI agent, automatically analyzes, categorizes, and summarizes failures from your monitors and synthetic checks. You can now receive a root cause analysis directly in your inbox, Slack, Teams, or any other alert channel the moment a check fails. This is live for all plans; enable it in the Rocky AI settings.

Playwright Check Suites improvements — permalink

Playwright Check Suites now detect pnpm and npm workspace structures and install the correct dependencies automatically, resolve package.json exports during bundling, and support .tsx and .jsx files. npx checkly pw-test is now GA with new --installCommand and --include flags, checks install all dependencies from your package.json (not just devDependencies), and the dependency cache refreshes automatically when your install command or working directory changes.

Dashboard revamp: compare checks more easily — permalink

Dashboard charts now display data based on the selected time range, so all checks on a dashboard share the same timeline, making it easier to compare results across services. Group names now appear next to check names where applicable, and you can save dashboard views based on your filters and search; saved views are personal and shareable via URL.

Checkly Agent: health probes and heartbeat monitoring — permalink

The Checkly Agent now exposes /-/readiness, /-/liveness, and /health endpoints on port 8080 for Kubernetes readiness and liveness probes. The agent also sends a heartbeat to the Checkly platform every 2 minutes with built-in retry logic, so Checkly can detect when a private location has no connected agents and notify your team via email, and a /metrics endpoint exposes agent-level Prometheus metrics like check slot utilization and job processing duration.

Workspace support for Playwright Check Suites — permalink

Playwright Check Suites now work with monorepos. If your project uses pnpm or npm workspaces, Checkly detects your workspace structure, uploads the relevant files, and installs the correct dependencies automatically, with no extra configuration. Shared utility packages are detected and included, workspace:* dependencies resolve correctly, and only the files and workspace packages your check suite needs are uploaded, not your entire repo.

Get started with the latest Checkly CLI — permalink

Workspace support is available starting with Checkly CLI v7.1. Update to the latest version, or run npm create checkly@latest if you are new to Checkly and prefer a guided setup.

npm install checkly@latest

Rocky AI root cause analysis for all check types — permalink

Rocky, Checkly's AI agent, can now analyze and triage failures for all uptime monitors and synthetic check types, including trace routes, packet captures, and other artifacts generated during a check run. You get error classification (code typo or infrastructure error), user impact (which user groups or features are affected), and root cause with supporting evidence and, where possible, a simple code fix.

Dev agent image for Private Locations — permalink

A new dev variant of the Private Location agent image is designed for checks that rely on npm packages with native dependencies requiring compilation. It works the same way as the standard agent but includes the build tools needed to compile native modules, and is available starting with version 6.3.0.

checkly/agent-dev:X.Y.Z

New role: Read & Run — permalink

Read & Run is a new role for users who need to run checks and tests without being able to change configuration. Members with this role can view all resources and manually trigger checks via the UI, API, or CLI, while deployments and config changes stay locked down.