Skip to content

About

Your whole Cloudflare Workers fleet on one page: 76 checks, free-limit warnings, an estimated bill, plain-word fixes and cleanup commands, several accounts in one install. Read-only. Free. No code changes.

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Repository files navigation

Ambersight logo: a magnifying glass with a hexagon lens

ambersight

Find the Cloudflare Worker that is about to break, run up a bill or sit exposed.
Your whole Workers fleet on one page. Read-only. Free. No code changes.

Deploy to Cloudflare

License: AGPL-3.0 Runs on the Workers Free plan Read-only API access

The Ambersight overview for one of several accounts: 6 of 12 Workers need attention, requests and errors today, storage limits and the estimated bill against a monthly budget

You have a dozen Workers, maybe fifty. One is quietly burning through the free plan's 100,000 daily requests. A cron job stopped running on Tuesday. A deploy last night doubled an error rate. An old Worker still answers on a public workers.dev address nobody remembers.

The Cloudflare dashboard shows you each Worker on its own. Ambersight shows you the fleet and tells you which Workers need attention, and why, in plain sentences.

It runs as one Worker in your own Cloudflare account. Every hour it reads your Workers' settings and metrics with a read-only API token, runs 83 checks, and gives you a ranked list of what to fix. One install can watch every account the token can see. Alerts go to e-mail, Slack, Discord, Telegram or any webhook.

  • On the Free plan, the risk is an outage. Past 100,000 requests in a day, every Worker in the account answers with Error 1027 until midnight UTC, and the same goes for 1,000 KV writes, 5 million D1 rows read and the other daily limits. Ambersight warns you at 70% and again at 90% of each one, with the time you will hit it at today's pace.
  • On Workers Paid, the risk is the bill. There is no hard spending cap, so a loop, a retry storm or a crawler is billed for every hour it runs. Ambersight compares each Worker's last hour with its usual hour and alerts you at 5x, for requests, CPU time, subrequests and D1, KV, R2, Queues and Durable Object operations. It also catches Durable Object or Queue work that runs while nobody uses the app, and shows the estimated bill in dollars.

New in 1.1.3: a security and privacy release. Tokens limited to chosen accounts, safer sign-in, client digests that hold only that client's data, and tools to stop recording or remove people's names. Read the upgrade notes before you deploy.

New in 1.1.2

  • A page for every Worker: requests, errors, CPU and wall time, limit hits and cron runs over 24 hours to 12 months, with every deploy marked and its before and after.
  • Analytics: trends against the period before, the bill per month, daily limits and storage growth with the date each cap is reached, time to resolve and which checks are only noise. Compare up to four Workers.
  • 13 months of history, kept as daily rollups. Every finding keeps its history: opened, alerted, acknowledged, resolved. A Findings inbox lists them by status, and acknowledge or snooze can be undone.
  • Smarter alerts: a nightly batch or a daily peak is judged against its own hour, and a spike no deploy explains names the change made just before it.
  • Monthly budget on Workers Paid, alerted within the hour, not a day later.
  • Read-only API tokens for Grafana or scripts (docs/api.md), CSV export on every chart and table, Ctrl+K to jump to any Worker, and an installable app that opens offline.
  • Fewer API calls: an hourly scan makes at most 3 GraphQL requests instead of up to 8, and pages are served from one cached row.
  • Explain in plain words on every finding: what it means, why it matters, and numbered steps with where to click, for people new to Cloudflare.
  • Clean up leftovers: unused KV, D1, R2, queues and idle Workers show the exact wrangler command to remove them (a backup first for D1). Ambersight still never deletes anything.
  • Export .md: save Needs attention or Housekeeping as a Markdown checklist, with every cleanup command in one block.
  • Safe to remove? Unused storage findings show when it was last used (from any source), its size and whether it is empty. Storage still in use from outside Workers gets no delete command.
  • Bound but idle: KV, D1, R2 and queues a Worker binds but has not used in 30 days.
  • Resources page: every KV, D1, R2 and queue with who binds it, size, last use and status, plus what deleting the leftovers would free.
  • Packages (opt-in): each direct npm dependency of a Worker as locked → latest, with a warning when one is a major version behind.

New in 0.1.1

  • Several accounts in one install, with an account switcher, an All accounts view, and a label, plan and client digest per account.
  • 74 checks, up from 39: every free daily limit (D1, KV, Durable Objects, Queues, Workflows, Workers AI), storage caps, idle loops, Pages, audit log, certificates, dangling DNS, stuck rollouts and config drift.
  • Estimated bill on Workers Paid: month to date, month-end forecast, the costliest Workers, and dollars on every runaway alert.
  • Richer timeline and digest: audit log, Cloudflare notifications and Pages deploys on the timeline; a monthly digest; a printable client report.
  • Lighter: an hourly scan of 200 Workers reads about 34,000 D1 rows instead of 293,000.

Everything is still read-only and free. The full list is in the changelog.

What it catches

Ambersight tells you when…
Free plan limits you are on track to hit today's 100,000-request limit (Error 1027), or any other daily limit: D1 rows, KV reads and writes, Durable Objects, Queues, Workflows, Workers AI. And roughly when.
Storage caps a D1 database nears 500 MB (Free) or its 10 GB hard cap (Paid), or KV, R2 or Durable Object storage nears the account limit.
Runaway usage requests, CPU time, subrequests, D1, KV, R2, Queues or Durable Objects jump to 5x their usual hourly level, or a brand-new Worker bursts. On Workers Paid there is no spending cap, so this is the alert that saves money.
Idle loops Durable Objects, a queue or a database stay busy for hours while the Workers using them get almost no requests: an alarm or retry loop.
Cost on Workers Paid: the estimated bill so far this month, the month-end forecast, the costliest Workers, and dollars on every runaway alert.
Errors and limits error rates pass 2% or 10%, or triple a Worker's own baseline; invocations hit CPU or memory limits; p99 CPU gets close to the free plan's 10 ms.
Cron health scheduled runs are missed or fail, a cron gets close to its CPU limit, or you are running out of Cron Triggers.
Bad deploys a problem starts right after a deploy, or a gradual rollout is left split for days. The timeline marks the deploy, and also shows audit log changes and Cloudflare's own notifications.
Config drift a binding or secret disappears after a deploy, staging and production differ, or a Worker nears 64 variables.
Pages a Pages deploy fails, or a project's pages.dev address skips the WAF of its custom domain.
Security a member or API token is added, a certificate is about to expire or stuck, or a DNS record points at a deleted Pages project anyone could claim.
Deprecations a Worker still uses Service Worker syntax, an old compatibility date, the legacy usage model, Workers Sites or redundant flags.
Exposure preview URLs are public, a forgotten workers.dev hostname still answers, or (optional) a URL is not behind Cloudflare Access.
Zombies and orphans Workers with no traffic or never deployed; KV, D1, R2 and Queues bound to nothing or bound but never used, with when each was last used; routes and service bindings that point at deleted Workers. Each one comes with the exact command or dashboard step to remove it.
Log costs a busy Worker logs every request at 100% sampling (logs are billed per GB from 1 December 2026), the Free plan's 0.5 GB a day will run out, Logpush is still on, or a D1 query reads millions of rows to return a few.
Packages (opt-in) a Worker's direct npm dependency is a major version behind the latest release, read from package-lock.json in its GitHub repository.
Ambersight itself its API token is about to expire, or scans keep failing.

The full list of checks has every threshold, ID and fix.

This is what a finding looks like (demo fleet):

Critical · auth-gateway
6.2% of requests failed in the last 24 hours. It started right after a deploy.
Compare the deploy of 2026-11-09 22:14 UTC with the one before it, or roll back. Open the Worker in Workers Observability and look at the most frequent exception.

Every finding comes with the evidence, a suggested fix, a link to the Cloudflare docs and an Explain in plain words section with step-by-step instructions. Findings from the same check are grouped, so fourteen Workers with public preview URLs are one row with one fix, not fourteen. Critical and warning come first; info-level housekeeping stays folded away. Acknowledge or snooze one finding or a whole group, or stop a check you don't care about. Export .md saves a group as a Markdown checklist.

Analytics and history. A page per Worker with traffic, errors, CPU and wall time, limit hits, cron runs and every deploy's before and after, over 24 hours to 12 months. An Analytics page with trends, the bill per month, daily limits and storage growth with the date each cap is reached, time to resolve and which checks are only noise. A findings inbox with each finding's history. Daily history is kept for 13 months, built from the data each scan already reads: no extra API calls. Everything exports as CSV, and a read-only API token serves it as JSON (docs/api.md). Each token reads only the accounts you choose for it, and who deployed or changed something only if you tick that too.

Also included: several accounts in one install with an "All accounts" view, groups and tags, per-Worker thresholds, include and exclude rules, a daily, weekly or monthly digest in your time zone (with a client digest per account), a printable client report, quiet hours, and a Self-cost panel that shows exactly what Ambersight itself uses.

Optional and free, each off until you uncomment it in wrangler.jsonc: a scan queue that scans every account on time, a plain-language digest summary from Workers AI, and Cloudflare Turnstile on the sign-in page. See configuration.md.

Packages (opt-in): map a Worker to its GitHub repository and Ambersight shows each direct npm dependency as locked version → latest, with a finding when one falls a major version behind. It needs a separate read-only GITHUB_TOKEN and is the only feature that reads your source code (package-lock.json only). See checks.md.

Screenshots

The page of one Worker, auth-gateway: requests, error rate, CPU per request and wall time over 7 days, with the deploy 14 hours ago marked where the error rate jumped to 6%
Every Worker on its own page. Traffic, errors, CPU and wall time over 24 hours to 12 months, with deploys marked.
Analytics over 30 days: requests, errors, error rate, CPU per request and limit hits with their change against the 30 days before, and the account's requests per day
Analytics. Trends against the period before, the Workers growing fastest, every Worker in one sortable table.
Analytics, Cost: the estimated bill per month, usage at list price per day, and Workers by cost per month and per million requests
Cost. On Workers Paid: the bill per month, spend per day and the Workers behind it.
Analytics, Limits and storage for a Free plan account: each daily limit per day against its free limit, with its peak and usual busiest hour, and a D1 database reaching its 500 MB cap around a date
Limits and storage. How close each daily limit runs, and the date each database reaches its cap.
Analytics, Reliability: findings opened and resolved per day, median time to resolve by severity, deploys per day, and alert quality per check with one check flagged as mostly silenced
Reliability. Time to resolve, and which checks are only noise to you.
Deploy timeline: deploys, a Pages deploy and audit log entries, with a deploy of auth-gateway flagged because the error rate rose right after it
Deploy timeline. Deploys, Pages builds and audit log changes. The deploy followed by a new problem is flagged.
The Findings inbox: open findings with acknowledge, snooze and history, and tabs for acknowledged, snoozed and resolved findings
Findings. Open, acknowledged, snoozed and resolved, each with its history.
All accounts: three Cloudflare accounts side by side, each with its critical, warning and info counts, Workers, requests today, error rate and free request budget
All accounts. One install watches every account the token can see.
Settings: scan frequency, time zone and history, and per account a label, client digest, plan and monthly budget
Settings. Accounts, budgets, checks, thresholds and alert routes, stored in D1. Export and import as JSON.
The overview in dark mode
Dark mode. Follows your system, or pick one.
The overview on a phone: a client account with a D1 database at 86% of its limit and half of its free requests used
On your phone. Install it as an app; it opens offline with data up to 7 days old.

Screenshots use a made-up demo fleet.

Why you can trust it with your account

  • Read-only, by design. It only sends GET requests and GraphQL queries to the Cloudflare API. The setup wizard reads your token's permissions and refuses to continue if it finds a write permission. If the token cannot show its permissions, you check it yourself and confirm. Every day Ambersight reads them again and raises a critical finding for anything it does not expect, or says that it could not check.
  • Your account, your data. It runs in your Cloudflare account and stores everything in your own D1 database. Nothing is sent to us. There is no telemetry and no tracking.
  • No code in your Workers. No SDK, no tail consumer, no redeploy. It reads what Cloudflare already records.
  • Free to run. With 30 Workers it uses well under 1% of the free plan's daily requests and 1 of your 5 Cron Triggers. Its own logs are off.
  • Open source. AGPL-3.0. Read every line before you deploy it.

Install

You need a Cloudflare account (the free plan is fine). Pick one way.

Best for Time
A. Deploy button Most people. No terminal. Updates deploy on every push to your copy. 5 minutes
B. Terminal Developers who use Wrangler and want the config in their own repo. 5 minutes

Both start with the same step.

Step 1: Create a read-only API token

Open the pre-filled token form. Check that every permission says Read, click Continue to summary, then Create Token. Copy the token now; Cloudflare shows it only once.

The form opens with the token set to All accounts and All zones. If you belong to several Cloudflare accounts, the setup wizard asks which ones to watch. You can also narrow the token to those accounts before you create it, which is worth doing: the token cannot change anything, but it can read a lot, including Worker code and stored data. See SECURITY.md.

Prefer to create the token by hand? These are the permissions.

My Profile › API Tokens › Create Token › Custom token, with only these permissions:

Scope Permission Access Needed for
Account Workers Scripts Read Required. The list of Workers and their settings
Account Account Analytics Read Required. Requests, errors, CPU, cron runs
Account Account Settings Read Finding your account ID
Account Workers KV Storage, D1, Workers R2 Storage, Queues Read Zombie and storage checks
Zone Workers Routes, Zone Read Exposure checks and orphaned routes
Account Cloudflare Pages Read Pages checks (failed deploys, pages.dev exposure, dangling DNS to Pages)
Zone SSL and Certificates, DNS Read Certificate and dangling-DNS checks
Account Notifications Read Optional: Cloudflare's own notifications on the timeline. Not in the pre-filled link (it has no short key); add it by hand.
Account Billing Read Detecting Free or Paid automatically. Without it, Ambersight uses PLAN from wrangler.jsonc; you can also set the plan in Settings › General.
Account Access: Apps and Policies Read Optional: the Access coverage check (off by default). Not in the pre-filled link; add it by hand if you turn the check on.
User API Tokens Read Letting the setup wizard and the daily check read the token's own permissions

Do not add any Edit permission. If the token can see several accounts, the setup wizard lets you tick the ones to watch. To fix the choice in config instead, set CF_ACCOUNT_ID in wrangler.jsonc under vars (comma-separated for several).

A. Deploy button

  1. Click Deploy to Cloudflare:

    Deploy to Cloudflare

  2. Sign in to Cloudflare and connect GitHub (or GitLab). Pick a name for your copy of the repository.

  3. Fill in the two fields:

    Field What to enter
    CF_API_TOKEN The read-only token from step 1
    ADMIN_PASSPHRASE A passphrase of at least 12 characters. You sign in with it, so put it in your password manager.
  4. Click Deploy. Cloudflare copies the repository into your account, creates the D1 database, applies the migrations and deploys the Worker.

  5. When it finishes, open the link on the deploy page (https://ambersight.<your-subdomain>.workers.dev), sign in, and follow the three setup steps. The first scan starts on its own and then runs every hour.

Your copy is connected to Workers Builds, so every commit you push to it deploys automatically. To change wrangler.jsonc (for example to turn on e-mail), edit the file on GitHub and commit. Whoever can push to that branch can run code with all of Ambersight's secrets, so protect it: see Updating.

B. Terminal

You need Node.js 22.12 or newer and git.

git clone https://github.com/jntegz/ambersight.git
cd ambersight
npm ci
npx wrangler login

# First deploy: builds the dashboard and creates the D1 database for you
npm run build
npx wrangler deploy

# Create the tables
npx wrangler d1 migrations apply DB --remote

# Add the two secrets (you will be asked to paste each value)
npx wrangler secret put CF_API_TOKEN
npx wrangler secret put ADMIN_PASSPHRASE

Wrangler prints your dashboard URL at the end of the deploy. Open it, sign in and follow the setup steps.

On the first deploy, Wrangler creates the ambersight D1 database and writes its ID into wrangler.jsonc. Commit that change if you keep your own repository. From then on, a single command builds, migrates and deploys:

npm run deploy

Updating

Ambersight shows a notice in the dashboard when a new release is out (it asks GitHub for the latest release number and sends nothing about your account; you can turn this off in Settings).

Update to a release you have looked at, not to whatever a branch holds at that moment. The name below is the release tag from the releases page.

  • Deploy button copy: in your repository, add this one as a remote and merge the release. Workers Builds deploys it.
    git remote add upstream https://github.com/jntegz/ambersight.git
    git fetch upstream --tags
    git merge v1.2.3
    git push
  • Terminal install: git fetch --tags && git merge v1.2.3, then npm ci && npm run deploy.

New database migrations are applied by the deploy script in both cases.

Protect the branch you deploy from: require a review and block force pushes (on GitHub, a branch ruleset). A malicious commit or dependency that reaches it runs inside the Worker with every secret and the whole database; SECURITY.md lists what that reaches. npm ci installs exactly what package-lock.json says.


Lock it down with Cloudflare Access

The dashboard has its own passphrase login, with hashed sessions, strict cookies and a lockout after wrong attempts (details in SECURITY.md). That is a good lock. Cloudflare Access is a better door: nobody reaches the sign-in page at all unless they first prove who they are with your e-mail, Google, GitHub or another identity provider. It also stops strangers from triggering the sign-in pause, which can keep a browser that has never signed in out for up to 15 minutes. Zero Trust is free for small teams.

Use both. It takes two minutes.

Option 1: Protect the whole Worker (recommended)

This covers every way to reach Ambersight: the workers.dev address, custom domains and routes.

  1. In the Cloudflare dashboard, go to Workers & Pages and open your ambersight Worker.
  2. Open the Access tab and choose Protect this Worker behind Access.
  3. Choose All traffic (not only previews) and click Apply Access.
  4. Edit the policy so it allows only you: for example Include › Emails › you@example.com. With the one-time PIN login method, Cloudflare e-mails you a code. No extra setup is needed.

Open your dashboard URL in a private window. You should see the Cloudflare Access login first, then the Ambersight passphrase.

Option 2: Custom domain with a self-hosted application

If you serve Ambersight on your own hostname, for example ambersight.example.com:

  1. Add the hostname under Workers & Pages › ambersight › Settings › Domains & Routes › Add › Custom domain.
  2. In Zero Trust › Access controls › Applications, choose Create new application › Self-hosted, add that public hostname, and attach a policy that allows only your e-mail address or group.
  3. Turn off the workers.dev address, or it stays a public way around Access. Set "workers_dev": false in wrangler.jsonc, commit it (for a Deploy-button copy, in your own repository) and deploy. Do not rely only on Settings › Domains & Routes › workers.dev › Disable in the dashboard: the next deploy applies wrangler.jsonc again and turns workers.dev back on.

Preview URLs are already off in Ambersight's config ("preview_urls": false), so there is no third way in.

Tip

Once Access is on, turn on the Cloudflare Access coverage check in Settings › Checks (it needs the Access: Apps and Policies Read permission). Ambersight will then tell you which of your other Workers are public without Access.


Alerts

E-mail (free)

Ambersight uses Cloudflare's send_email binding. It is free when you send to a verified address in Email Routing; a Gmail address works.

  1. Turn on Email Routing for a domain you have on Cloudflare and verify the address you want alerts at.
  2. In wrangler.jsonc, set EMAIL_FROM (any address on that domain) and DIGEST_TO (your verified address), and uncomment the send_email block with the same address.
  3. Commit (Deploy button) or run npm run deploy (terminal).

Slack, Discord, Telegram, ntfy or any webhook

npx wrangler secret put WEBHOOK_URL

Paste a Slack or Discord incoming-webhook URL, a Telegram https://api.telegram.org/bot<token>/sendMessage?chat_id=<id> URL, an ntfy topic URL, or any HTTPS endpoint that accepts JSON. With the Deploy button you can also add it in the Cloudflare dashboard under ambersight › Settings › Variables and Secrets.

In Settings › Alerts, choose per severity whether alerts go to e-mail, the webhook, both or only the digest. Every channel has a Send test button.

docs/alerts.md covers the defaults, quiet hours, the digest, per-group digests and the exact message format for each channel.

What it costs

Nothing, on the Workers Free plan. With 30 Workers:

  • Cron Triggers: 1 of your 5. It fires every 15 minutes but only scans at the interval you choose (hourly by default); the other runs stop after a few small D1 queries.
  • Requests: about 100 cron runs plus a few hundred dashboard API calls a day. That is well under 1% of the 100,000 daily requests. Page views are static assets and do not count.
  • Subrequests: at most 42 per scan, within the free limit of 50. A typical hourly scan uses fewer than 10.
  • D1 and CPU: about 10 D1 calls per scan and a small fraction of the free allowance. Its own logs are off.
  • Optional extras: the scan queue and the Workers AI summary each stay well inside their own free allowances; see configuration.md.

The Self-cost panel in Settings shows the measured numbers for your account.


FAQ

Can Ambersight change or delete anything in my account?

No. It only sends GET requests and GraphQL queries, and the token you give it should have no write permissions. The setup wizard checks the token's permissions and refuses to continue if it finds a write permission; if the token cannot show them, you confirm that yourself. A daily check raises a critical finding if the token gains a permission Ambersight does not expect. Even if the code had a bug, a read-only token cannot change anything.

What happens if someone steals the API token?

They could read your Workers' names, settings and metrics. They could not deploy, edit or delete anything. The token is stored as a Worker secret: it is never sent to the browser, logged or written to D1. You can roll or delete it any time under My Profile › API Tokens.

Do I need to change my Workers or add an SDK?

No. Ambersight reads what Cloudflare already records: the Workers API, the GraphQL Analytics API and your account's resource lists. Your Workers are never touched or redeployed.

Does any of my data leave my Cloudflare account?

No, apart from what you set up yourself. Ambersight talks only to api.cloudflare.com and, for the optional update notice, to api.github.com to read the latest release number (no account data is sent). Two more hosts are contacted only if you turn them on: Cloudflare's public status page www.cloudflarestatus.com (the incidents check, off by default) and challenges.cloudflare.com (Turnstile on the sign-in page). Alerts go only to the e-mail address and webhook you configure. There is no telemetry.

How is this different from the Cloudflare dashboard or Workers Observability?

The dashboard and Observability are great at showing one Worker in depth. Ambersight looks across all of them and answers a different question: which ones need my attention right now? It compares every Worker with its own history, puts deploys next to the problems that followed them, and checks things the dashboard does not flag, such as forgotten workers.dev hostnames, unbound KV namespaces or Workers that have not had a request in a month. Use it to find the Worker, then open Observability to dig in.

I'm on Workers Paid. Is it still useful?

Yes, maybe more so. Workers Paid has no hard spending cap, so a loop, a retry storm or a crawler can run up a bill. The runaway checks compare every Worker's last hour with its usual level and alert you when requests, CPU time, subrequests or storage operations jump. The plan is detected automatically, and the setup wizard tells you if it could not be. The free-budget meter is hidden on Paid.

Will it fit in the free plan next to my other Workers?

Yes. It uses 1 Cron Trigger and well under 1% of the daily request allowance for a fleet of 30 Workers. See What it costs. The Self-cost panel shows the real numbers.

Some checks say a dataset is unavailable. Is something broken?

Probably not. A few Cloudflare analytics datasets (for example scheduled invocations and some storage analytics) are not available on every plan or account. Ambersight tries each one, uses what it gets, and tells you what it had to skip. The other checks keep working. Troubleshooting explains each message.

Can I monitor several Cloudflare accounts?

Yes. If the token can see several accounts, tick the ones to watch in the setup wizard or Settings › Accounts. Read-only API tokens can be limited to some of them. Dashboard sessions cannot: there is one shared passphrase, which cannot identify separate people, so everyone signed in sees every watched account. The header gets an account switcher and an All accounts view, each account can have its own label and plan, and each can get its own digest by e-mail (handy for clients). Each cron run scans the most overdue account, so with 8 accounts each is scanned about every 2 hours. Add the free scan queue to scan them all on time (configuration.md).

I forgot my passphrase.

Set a new one with npx wrangler secret put ADMIN_PASSPHRASE (or in the dashboard under Variables and Secrets). Use a value different from the current secret, then sign in with it.

This also works if you changed the passphrase in Settings › Security: the one stored there (hashed, in D1) counts only while the secret stays the same. The exception is a Settings passphrase that has not been used to sign in since you upgraded to this version, or an install that had no secret. Then remove the stored one, and the ADMIN_PASSPHRASE secret applies again:

npx wrangler d1 execute ambersight --remote --command "DELETE FROM settings WHERE key IN ('passphrase_hash', 'passphrase_binding')"
I think someone else has my passphrase or a session.

Replace the secret with a new value, as above. That ends every session, every remembered browser and every read-only API token issued under the old credential, whether or not a passphrase was set in Settings. If you replace it right after upgrading, before anyone has signed in and before the first scheduled run, also revoke tokens created before the upgrade: they cannot be told apart from new ones yet. If a Settings passphrase has not been used since you upgraded, also run the DELETE command above, because it has nothing tying it to the secret yet.

Then sign in with the new secret, create new API tokens where you need them and revoke the old ones in Settings › Security. If the Cloudflare API token may be exposed too, roll it under My Profile › API Tokens.

Can I tune the checks for my fleet?

Yes. Every check can be switched off, and every threshold can be changed globally, per group or per Worker in Settings. Group Workers by name pattern (Cloudflare tags become groups automatically), ignore the ones you do not care about, and export the whole setup as config.json to keep it in version control. Every field is listed in docs/configuration.md.

The license is AGPL. Can I use it at work?

Yes. Running Ambersight for your own company or team, modified or not, is fine. The AGPL asks one thing more than the GPL: if you modify Ambersight and let other people use your modified version over a network (for example as a hosted service for customers), you must offer them your source code under the same license. The name and logo are covered separately; see TRADEMARK.md.

Is Ambersight made by Cloudflare?

No. It is an independent open-source project built on Cloudflare's public APIs. It is not affiliated with or endorsed by Cloudflare.


Documentation

  • Checks: all 83 checks, with thresholds, IDs and fixes
  • How it works: what one scan does, what it stores and how it stays inside the free plan
  • Alerts: channels, defaults, the digest and message formats
  • Configuration: wrangler.jsonc, secrets and every config.json field
  • Troubleshooting: scan warnings, token errors, e-mail that does not arrive
  • Changelog

Develop

npm install
cp .dev.vars.example .dev.vars   # a read-only token and a passphrase
npm run db:migrate:local
npm run dev                       # builds the UI and starts wrangler dev on :8787
npm run dev:web                   # optional: Vite with hot reload, proxying /api to :8787
npm test && npm run typecheck
Path What it is
src/worker/ The Worker: API (index.ts), scan (scan.ts), checks (checks.ts, hygiene.ts), Cloudflare client (cf.ts), setup wizard, auth, alerts (notify.ts)
src/shared/ Types and the config model shared by the Worker and the UI
web/ Svelte 5 dashboard, built by Vite into dist/ and served as static assets
migrations/ D1 schema
test/ Vitest: unit tests, plus end-to-end scans against fake API responses with D1 emulated on node:sqlite

Contributions are welcome. Please read CONTRIBUTING.md first; the short version is read-only always, free plan first, tests for every check. To report a security issue, see SECURITY.md.

License

Ambersight is free software under the GNU Affero General Public License v3.0. The name Ambersight and the logo are not covered by the license; see TRADEMARK.md.

Ambersight is an independent project and is not affiliated with or endorsed by Cloudflare, Inc. "Cloudflare" and "Workers" are trademarks of Cloudflare, Inc., used here only to describe compatibility.

Created by Jan Tegze

About

Your whole Cloudflare Workers fleet on one page: 76 checks, free-limit warnings, an estimated bill, plain-word fixes and cleanup commands, several accounts in one install. Read-only. Free. No code changes.

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages