Cloudflare Service Keys: Audit Old Automation After September 30

Illustrated infographic summarizing: Cloudflare Service Keys: Audit Old Automation Before September 30

By Greg Nowak. Updated November 16, 2026.

Cloudflare lists September 30, 2026 as the end-of-life date for Service Key authentication. That deadline has passed. If your team never completed the migration, treat automation using X-Auth-User-Service-Key as an unsupported dependency that needs attention.

A working website does not prove the cleanup is finished. A certificate job, deployment helper, or recovery script may sit unused until someone needs it. For business owners and operations leads, the useful question is: which business processes still depend on these credentials, and who can demonstrate that their replacements work?

Start with the jobs that affect your next deployment, certificate operation, or recovery procedure. Give each an owner, replace its authentication, and test it from the environment where it actually runs.

What changed, and which credentials matter?

Cloudflare’s API deprecation notice identifies requests authenticated through the Service Key header and names API Tokens as the replacement. It also calls out older cloudflared installations and origin-ca-issuer versions that need API Token support.

Origin CA keys belong in this audit. They begin with v1.0- and have access to every account their user can access. For Origin CA certificate operations, Cloudflare specifies an API token with Zone → SSL and Certificates → Edit permissions.

Classify each credential before changing it. Finding “Cloudflare” in a DNS script does not establish that it uses Service Key authentication. Record the actual header or authentication method so the team can distinguish affected jobs from unrelated maintenance.

Find the callers, including the forgotten ones

Search repositories, deployment directories, infrastructure configuration, and scheduled tasks. These commands list matching filenames without printing potentially sensitive matching lines:

rg -l --hidden -g '!.git' 'X-Auth-User-Service-Key|v1\.0-' .
rg -l --hidden -g '!.git' 'origin-ca|/certificates|cloudflare' .

The second search deliberately casts a wider net. Review matches locally: a filename is a lead, not evidence that a job is active. Ripgrep normally respects ignore files, so separately inspect deployment files or secret locations that your repository search excludes.

Check CI variables, hosting dashboards, secret managers, cron entries, and third-party integration settings too. A credential injected at runtime may never appear in source control.

For each active caller, record its purpose, account and zones, execution environment, schedule, secret location, and responsible person. Include the last successful run and the consequence of failure. Agencies should keep that inventory separated by client, with a clear contact for approving changes.

Prioritise by business consequence

If a job is already failing, identify its authentication method and capture the error before changing several things at once. Then work through dependencies according to their next required use.

Dependency Audit priority Evidence needed to close it
Failed production job Immediate investigation Successful operation and confirmed intended result
Certificate issuance or deployment helper Before its next required run Controlled test through the actual tool and runner
Recovery or occasional maintenance script Schedule an explicit test Rehearsal against a safe target
Apparently abandoned integration Confirm ownership and usage Documented retirement decision and secret cleanup
Use business impact to order the audit. Infrequent jobs need deliberate testing because routine monitoring may never exercise them.

Choose a token the team can maintain

Use one token per automation as a practical default. Give it a descriptive name, such as the client, environment, and job. This makes later rotation and incident investigation easier.

Match permissions to the operation and resources to the intended zones. A DNS updater may need DNS Edit; an inventory job may only need DNS Read. Avoid granting account-wide access merely to get past a failed test. Cloudflare’s token creation guide explains resource policies, optional expiration, and IP filtering.

For durable shared automation, consider an account API token. It operates independently of an employee’s user identity. Check endpoint compatibility first. Cloudflare’s account token documentation now permits creation by members with API Token Provisioning capabilities or Super Administrator status; members can only grant permissions within their own access.

Store the secret immediately in the appropriate vault or CI secret store: Cloudflare displays it only once. Set expiration alongside a realistic rotation owner and reminder. Use IP restrictions where the runner’s outbound addresses are predictable.

Test authentication, permissions, and the actual job

Have your secret store inject CLOUDFLARE_API_TOKEN into the test environment. For a user token, check its status with:

curl "https://api.cloudflare.com/client/v4/user/tokens/verify" \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN"

For an account token, set ACCOUNT_ID and use the account verification endpoint:

curl "https://api.cloudflare.com/client/v4/accounts/$ACCOUNT_ID/tokens/verify" \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN"

An active response establishes token status. Follow it with the operation the job needs: read the intended resource, or make a controlled write with a documented reversal. Where practical, also confirm that an unrelated resource is inaccessible.

Update the client to send Authorization: Bearer and remove the legacy Service Key header. Run it through the real scheduler or deployment system to check secret injection and network restrictions. After the deadline, a fallback plan must use supported authentication or an agreed manual procedure.

Close the audit with a usable handover

Record the tested version, token scope, secret location, owner, rotation procedure, and failure alert destination. Remove retired credentials from runtime configuration and documentation after accounting for their callers. Check the next normal run as well as the controlled test.

If responsibility is split across client accounts, suppliers, and older tooling, I can help turn the inventory into a tested migration and clear handover. Discuss your Cloudflare automation audit with Greg.

Related on GrN.dk

Need help with this kind of work?

Discuss your Cloudflare automation audit with Greg Get in touch with Greg.

Sources

Latest articles

WordPress jobs running late? Learn how to separate scheduling problems, task errors and a growing backlog, then choose the right scheduler or runner.

Before changing an AI workflow, check that it still does the job. Use realistic examples, clear release criteria and a comparison with the version running today.

Sending an invoice does not mean it has arrived. Understand delivery statuses, the system requirements coming in 2027, and how focused AI support can help finance teams handle errors and follow-up.

Before scaling AI content, check the pages you already have. Verify claims, resolve overlapping coverage, and give WordPress reviewers control over publication.

Build a refund approval workflow in n8n and Stripe that shows reviewers the exact action, expires old approvals and tracks refunds through to completion.

A timeout can hide a successful job creation. Use stable operation IDs, destination API rules and a clear recovery process to prevent duplicate work.

Does your ChatGPT subscription cover website translation? Understand API credits, access and a WordPress workflow that turns approved Danish content into translation drafts.

An internal AI can keep quoting a deleted policy. Check document updates, indexed passages and cached answers to find where outdated guidance survives.

Codex drew all ten pages of my Ninjai comic from the script Claude wrote in the first test. What made that handover work, and what it cost.

ChatGPT delivered ten painted comic pages; Claude drew one page in code. A practical look at AI tool selection, creative briefs and approval decisions.