<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>RADAR — Cloudflare developer changelog</title>
    <link>https://radar.trulysovereign.com</link>
    <atom:link href="https://radar.trulysovereign.com/feeds/source/cloudflare-developer-changelog.xml" rel="self" type="application/rss+xml"/>
    <description>Changes RADAR detected at Cloudflare developer changelog.</description>
    <language>en</language>
    <lastBuildDate>Sat, 21 Nov 2026 19:19:18 GMT</lastBuildDate>
    <item>
      <title>Agents SDK v0.34.0 upgrades PiHarness to Pi 1.6, breaking backward compatibility</title>
      <link>https://radar.trulysovereign.com/items/85b62f74-3876-4f44-b50b-a9014d15a7e7</link>
      <guid isPermaLink="false">85b62f74-3876-4f44-b50b-a9014d15a7e7</guid>
      <pubDate>Sat, 21 Nov 2026 19:18:45 GMT</pubDate>
      <category>migrate</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-20-pi-harness-pi-1-6/">Cloudflare developer changelog</source>
      <description>Cloudflare Agents SDK v0.34.0 upgrades PiHarness from Pi 1.5 to Pi 1.6. Once a conversation runs on Pi 1.6, Pi 1.5 cannot run its tool calls or model requests. API method signatures changed: harness.signal() replaces (await harness.pi()).signal(), and harness.root() must be called before reaching the root session through harness.pi(). A new error type MCPAuthorizationRequiredError is introduced for MCP re-authorization. The claudeCode() default model changed to claude-sonnet-5-5. Workers AI models were retired on November 20, 2026.

Why it matters: Rolling back to Pi 1.5 after upgrading will break existing conversations that have run on Pi 1.6. Applications using the old signal delivery pattern or accessing the root session without calling harness.root() first will fail. MCP client implementations that do not catch the new error type will not handle re-authorization correctly.

What to do: Review the Pi Durable 1.6.0 changelog and Agents SDK v0.34.0 changelog for the full list of breaking changes. Update signal delivery calls from (await harness.pi()).signal() to harness.signal(). Add harness.root() calls before accessing the root session. Implement MCPAuthorizationRequiredError handling to redirect users to error.authUrl and provide a way to call reconnectMcpServer(id). Plan to roll forward rather than roll back if deployment issues occur.</description>
    </item>
    <item>
      <title>Cloudflare Browser Run adds /script endpoint for single-request automation</title>
      <link>https://radar.trulysovereign.com/items/d2bf6b39-e650-4d44-a4c4-dcad3eb8d8d2</link>
      <guid isPermaLink="false">d2bf6b39-e650-4d44-a4c4-dcad3eb8d8d2</guid>
      <pubDate>Sat, 21 Nov 2026 07:10:33 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-20-script-endpoint/">Cloudflare developer changelog</source>
      <description>Cloudflare Browser Run introduced POST /script and POST /script/&lt;sessionId&gt; endpoints that execute Playwright, Puppeteer, or Chrome DevTools Protocol functions in a single API request without deploying a Worker. The endpoints return session and page identifiers in cf-browser-session-id and cf-browser-target-id response headers. Sessions remain open after script execution until explicitly closed with DELETE /devtools/browser/&lt;sessionId&gt; or until idle timeout. Workers can call the same endpoints through env.BROWSER.script() without an API token. A Script Studio interface in the Cloudflare dashboard allows writing and running scripts against live browsers without an API token.

Why it matters: Existing Browser Run workflows that deploy Workers for automation can now run the same code in a single API call, eliminating the deployment step. Sessions that are not explicitly closed will consume resources until idle timeout, changing the cleanup model from previous Browser Run usage patterns.

What to do: Review current Browser Run Worker deployments to identify automation that can move to /script endpoints. Add explicit session cleanup with DELETE /devtools/browser/&lt;sessionId&gt; or browser.close() calls to any new /script usage to avoid sessions running to idle timeout.</description>
    </item>
    <item>
      <title>Cloudflare blocks dispatch namespace renames via API and CLI</title>
      <link>https://radar.trulysovereign.com/items/eb1d7879-b34d-4755-b886-6ac22defd53f</link>
      <guid isPermaLink="false">eb1d7879-b34d-4755-b886-6ac22defd53f</guid>
      <pubDate>Sat, 21 Nov 2026 07:10:19 GMT</pubDate>
      <category>migrate</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-20-dispatch-namespace-rename-removed/">Cloudflare developer changelog</source>
      <description>PUT and PATCH requests to the dispatch namespace API now return a 400 error with code 100413 if they attempt to change the name field. The wrangler dispatch-namespace rename command and Terraform cloudflare_workers_for_platforms_dispatch_namespace resource name changes fail with the same error. Cloudflare states this prevents failures where dispatch Workers and outbound Workers retained the old namespace name after a rename, causing subsequent API requests to read, deploy, or delete Workers to fail.

Why it matters: Any automation or infrastructure-as-code that renames dispatch namespaces will break. Terraform applies that change the name attribute will fail. Existing runbooks or scripts that use wrangler dispatch-namespace rename will stop working.

What to do: Audit your Terraform configurations, CI/CD pipelines, and operational scripts for dispatch namespace renames. Replace rename operations with the documented five-step migration: create new namespace, deploy user Workers to it, update dispatch Worker binding, verify routing, delete old namespace. Update any runbooks that reference wrangler dispatch-namespace rename.</description>
    </item>
    <item>
      <title>Hyperdrive supports custom CA and mTLS over Workers VPC</title>
      <link>https://radar.trulysovereign.com/items/f2ab74af-5a12-428f-8a13-1d79c80fc193</link>
      <guid isPermaLink="false">f2ab74af-5a12-428f-8a13-1d79c80fc193</guid>
      <pubDate>Sat, 21 Nov 2026 07:10:06 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-18-hyperdrive-vpc-custom-ca-mtls/">Cloudflare developer changelog</source>
      <description>Hyperdrive can now connect to private databases over Workers VPC using custom certificate authority (CA) certificates and client certificate authentication (mTLS). This covers PostgreSQL databases and Amazon RDS instances with private CA or RDS CA bundle certificates, and client certificates with RSA or EC P-256 keys. Certificates attach to the VPC Service, eliminating the previous requirement to route traffic through Cloudflare Access and Cloudflare Tunnel. Configuration requires Wrangler 4.164.0 or later, or Terraform provider 5.29.0 or later.

Why it matters: Teams connecting Hyperdrive to databases with custom certificates can remove the Access application, service token, and WebSocket handling previously required for this setup. The direct VPC connection simplifies the architecture and removes the Access layer from the database path.

What to do: If you currently route Hyperdrive database traffic through Cloudflare Access and Tunnel to handle custom certificates, evaluate migrating to the VPC Service attachment method using the wrangler cert upload and wrangler vpc service create commands shown in the changelog.</description>
    </item>
    <item>
      <title>CVE-2025-6087 SSRF vulnerability fixed in @opennextjs/cloudflare v1.3.0</title>
      <link>https://radar.trulysovereign.com/items/c694d3bc-ea3f-4004-97cd-cf355b5e8d12</link>
      <guid isPermaLink="false">c694d3bc-ea3f-4004-97cd-cf355b5e8d12</guid>
      <pubDate>Fri, 20 Nov 2026 20:24:16 GMT</pubDate>
      <category>migrate</category>
      <source url="https://developers.cloudflare.com/changelog/post/2025-06-05-open-next-size/">Cloudflare developer changelog</source>
      <description>Cloudflare published CVE-2025-6087 for an SSRF vulnerability in the @opennextjs/cloudflare package. The vulnerability is fixed in v1.3.0 and later versions.

Why it matters: SSRF vulnerabilities allow attackers to make the server perform requests to unintended destinations, potentially accessing internal resources or services that should not be externally reachable.

What to do: Update @opennextjs/cloudflare to v1.3.0 or later. Check your package.json and lock files for any version below v1.3.0.</description>
    </item>
    <item>
      <title>Cloudflare WAF adds CVE-specific rules for PHP and Ruby RCE vulnerabilities</title>
      <link>https://radar.trulysovereign.com/items/3f30ca3c-e116-498e-97d2-42a7290a4a99</link>
      <guid isPermaLink="false">3f30ca3c-e116-498e-97d2-42a7290a4a99</guid>
      <pubDate>Fri, 20 Nov 2026 07:19:48 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-19-emergency-waf-release/">Cloudflare developer changelog</source>
      <description>Cloudflare deployed WAF rules targeting CVE-2026-109618, CVE-2026-109619, CVE-2026-115560, CVE-2026-126714 (PHP RCE), CVE-2026-13282 (Ruby heap overflow), GHSA-87fm-7w84-j2px (Ruby integer overflow), CVE-2026-129289 (PHP-FPM), CVE-2026-15353 and CVE-2026-129506 (PHP SOAP). The PHP RCE rule detection was updated. New PHP-FPM and PHP SOAP rules default to log mode; all other new rules default to block mode.

Why it matters: Applications running vulnerable PHP or Ruby versions behind Cloudflare WAF are now protected by default against these remote code execution and overflow vulnerabilities. The PHP-FPM and SOAP rules log rather than block to reduce false positive risk, requiring manual review before enforcement.

What to do: Check Cloudflare WAF logs for requests matching the new PHP-FPM (CVE-2026-129289) and PHP SOAP (CVE-2026-15353, CVE-2026-129506) rules. Override to block mode after confirming no false positives. Verify applications are not vulnerable to the listed CVEs.</description>
    </item>
    <item>
      <title>Cloudflare Images adds Gaussian sharpening mode with configurable parameters</title>
      <link>https://radar.trulysovereign.com/items/cafeb085-9e38-49e6-8c4a-31223ec35d45</link>
      <guid isPermaLink="false">cafeb085-9e38-49e6-8c4a-31223ec35d45</guid>
      <pubDate>Fri, 20 Nov 2026 07:19:33 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-19-configurable-sharpening/">Cloudflare developer changelog</source>
      <description>Cloudflare Images sharpen parameter now accepts a structured object with mode=gaussian alongside amount (0-10), sigma (0-5), and optional threshold (0-255) settings. Existing scalar values like sharpen=5 continue using the box filter. The structured format works in URL parameters, Workers fetch() cf.image object, and Images binding .transform() calls.

Why it matters: Gaussian sharpening provides finer control over edge detection for photography and complex images, reducing unwanted sharpening of noise or grain that box filtering treats uniformly. Different content types—text, product photos, illustrations—can now use tailored sharpening configurations instead of a single intensity value.

What to do: Review images currently using scalar sharpen values. For photography or content where box sharpening over-emphasizes grain, test Gaussian mode with amount and sigma settings appropriate to the content type. Update transform calls to use the structured object format if switching modes.</description>
    </item>
    <item>
      <title>Cloudflare deletes Worker versions and Pages deployments after 30 days</title>
      <link>https://radar.trulysovereign.com/items/d712a63d-baa1-4ed6-96c2-c8f095fb892e</link>
      <guid isPermaLink="false">d712a63d-baa1-4ed6-96c2-c8f095fb892e</guid>
      <pubDate>Thu, 19 Nov 2026 20:40:26 GMT</pubDate>
      <category>migrate</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-19-workers-pages-version-retention/">Cloudflare developer changelog</source>
      <description>Cloudflare now deletes Worker versions and Pages deployments older than 30 days, except the 100 most recent versions and currently deployed versions. Deleted versions no longer appear in the dashboard or API, and their Version URLs and preview URLs stop working. Effective 19 November 2026.

Why it matters: Any automation, documentation, or testing that relies on Version URLs or Pages preview URLs older than 30 days will break. Rollback capability is limited to the 100 most recent versions instead of full history.

What to do: Audit CI/CD pipelines, documentation, and test suites for hardcoded Version URLs or Pages preview URLs. Replace Version URLs with Worker Previews for branch and pull request workflows, which always serve the latest deployment with production bindings.</description>
    </item>
    <item>
      <title>Browser Run /snapshot endpoint adds PDF, links, scrape, and JSON formats</title>
      <link>https://radar.trulysovereign.com/items/b983b372-a333-4f35-931d-b2a57b79f7ff</link>
      <guid isPermaLink="false">b983b372-a333-4f35-931d-b2a57b79f7ff</guid>
      <pubDate>Thu, 19 Nov 2026 20:40:18 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-19-snapshot-all-formats/">Cloudflare developer changelog</source>
      <description>Cloudflare Browser Run's /snapshot endpoint now accepts four additional format values: pdf, links, scrape, and json. Previously /snapshot returned only HTML, screenshots, Markdown, and accessibility tree. The json format incurs Workers AI usage charges. Formats are captured in a fixed order regardless of array order, with PDF printed last. The actionTimeout budget (60s default, 120s max) now covers navigation, captures, retries, and AI requests. Requests must specify at least two distinct formats.

Why it matters: Workflows that previously required multiple page loads to capture different formats (such as a PDF plus structured data) can now complete in a single request, reducing latency and Browser Run invocations. The json format adds Workers AI costs that were not present before. The fixed capture order and shared timeout budget change how multi-format requests behave under time pressure.

What to do: Review existing /snapshot calls. If you currently make separate requests for PDFs, links, scraped elements, or JSON extraction from the same page, consolidate them into one /snapshot call with multiple formats. For json format requests, add prompt or response_format as required and account for Workers AI usage in your budget. Verify that your actionTimeout settings accommodate the combined work of all requested formats.</description>
    </item>
    <item>
      <title>Stream emits ERR_LIVE_INPUT_LLHLS_INCOMPATIBLE when B-frames detected in LL-HLS</title>
      <link>https://radar.trulysovereign.com/items/cb54cc52-b736-4d14-b0d8-e0f36ba25694</link>
      <guid isPermaLink="false">cb54cc52-b736-4d14-b0d8-e0f36ba25694</guid>
      <pubDate>Thu, 19 Nov 2026 20:40:03 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-19-ll-hls-b-frame-warning/">Cloudflare developer changelog</source>
      <description>Cloudflare Stream now emits warning code ERR_LIVE_INPUT_LLHLS_INCOMPATIBLE once per live session when B-frames are detected in Low-Latency HLS broadcasts. The warning appears in the live input's Logs tab, live webhooks, and GraphQL input events. B-frames are not supported when LL-HLS is enabled for a live input.

Why it matters: B-frames in LL-HLS streams cause freezing and stuttering because they reference content not yet transmitted, stalling delivery preparation. The new warning surfaces a configuration error that degrades viewer experience but may have gone unnoticed without explicit detection.

What to do: If you use Cloudflare Stream with Low-Latency HLS enabled, check your encoder settings and disable B-frames in your broadcast application. Alternatively, disable Low-Latency HLS in your live input settings if higher playback latency is acceptable.</description>
    </item>
    <item>
      <title>Workers VPC binding cf1:network now egresses through Gateway by default</title>
      <link>https://radar.trulysovereign.com/items/c3f68b5d-1231-46e2-88fb-1bd5d7fe240d</link>
      <guid isPermaLink="false">c3f68b5d-1231-46e2-88fb-1bd5d7fe240d</guid>
      <pubDate>Thu, 19 Nov 2026 20:39:55 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-06-05-gateway-egress/">Cloudflare developer changelog</source>
      <description>Workers using VPC Network binding with network_id: &quot;cf1:network&quot; now route public Internet egress through Cloudflare Gateway automatically. Previously this required Cloudflare Tunnel, Mesh, or WAN on-ramp; those are no longer required. Zero Trust traffic policies (DNS, HTTP, Network, and egress) now apply to Worker-originated traffic the same way they apply to WARP users.

Why it matters: Existing Zero Trust policies that were scoped to WARP users now also enforce on Worker egress traffic. If those policies block or filter traffic your Workers depend on, those Workers will fail. Gateway logs will now capture all Worker egress, changing what appears in audit trails and potentially what triggers alerts.

What to do: Review your Zero Trust DNS, HTTP, Network and egress policies in the Gateway dashboard. Verify none of them block hostnames or IP ranges your Workers with cf1:network bindings call. Check Gateway logs for new Worker egress entries and confirm they match expected behavior.</description>
    </item>
    <item>
      <title>Cloudflare Access adds automatic SAML IdP certificate retrieval</title>
      <link>https://radar.trulysovereign.com/items/a7ccb84d-79ec-4643-baa3-7d7529becdf5</link>
      <guid isPermaLink="false">a7ccb84d-79ec-4643-baa3-7d7529becdf5</guid>
      <pubDate>Wed, 18 Nov 2026 20:36:19 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-18-saml-idp-certificate-discovery/">Cloudflare developer changelog</source>
      <description>Cloudflare Access can now retrieve SAML signing certificates from an identity provider's metadata endpoint. In the generic SAML configuration, set Signing certificate source to &quot;Fetch certificates from IdP metadata&quot; and provide the IdP metadata URL. For API configurations, use the metadata_url field instead of manual certificate fields. Access caches the certificates and does not poll the endpoint. Removing a compromised certificate from metadata does not immediately revoke the cached certificate or existing Access sessions.

Why it matters: Automatic certificate retrieval eliminates manual certificate uploads when IdPs rotate signing certificates, but introduces a revocation delay risk. A compromised certificate removed from IdP metadata remains valid in Access's cache and does not terminate existing sessions, creating a window where revoked credentials continue to authenticate.

What to do: If you use Cloudflare Access with SAML authentication, review the setup and emergency replacement guidance at the linked documentation to understand the revocation delay behavior before enabling automatic certificate discovery. For existing manual certificate configurations, no action is required unless you choose to migrate to the metadata_url approach.</description>
    </item>
    <item>
      <title>Cloudflare WAF adds 6 Java4Shell CVE-2026-114938 detections in log mode</title>
      <link>https://radar.trulysovereign.com/items/636513f6-7b58-4813-b0ec-d684c101aa6f</link>
      <guid isPermaLink="false">636513f6-7b58-4813-b0ec-d684c101aa6f</guid>
      <pubDate>Wed, 18 Nov 2026 20:36:09 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-18-emergency-waf-release/">Cloudflare developer changelog</source>
      <description>Cloudflare deployed 6 new WAF rules targeting malformed base64 patterns associated with CVE-2026-114938 (Java4Shell). The rules detect malformed base64 in headers, query values, body values, path segments, JWT components, and WebAuthn values. All rules are deployed in log mode by default, not blocking mode.

Why it matters: Applications behind Cloudflare WAF will log but not block requests matching these patterns. If CVE-2026-114938 affects your Java components, you are collecting detection data but not preventing exploitation until you override the default action to block.

What to do: Review WAF logs for rule IDs 4d718bf4, 2337b51d, e5e5836c, 6fd6f152, 5a25d991, 67d74ebbN. If your application uses Java components vulnerable to CVE-2026-114938, override these rules to block mode following Cloudflare's blog post guidance.</description>
    </item>
    <item>
      <title>Workers VPC adds custom CA and mTLS client certificate support</title>
      <link>https://radar.trulysovereign.com/items/14cae17a-9508-4078-8db2-d741a7361232</link>
      <guid isPermaLink="false">14cae17a-9508-4078-8db2-d741a7361232</guid>
      <pubDate>Wed, 18 Nov 2026 07:32:07 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-18-custom-ca-mtls/">Cloudflare developer changelog</source>
      <description>Cloudflare Workers VPC Services now accept custom certificate authority (CA) certificates and client certificates for mutual TLS. Previously, Workers VPC trusted only publicly trusted CAs and Cloudflare Origin CA, requiring verify_ca or disabled modes to connect to private CAs, and could not reach origins requiring client certificates. Configuration requires Wrangler 4.164.0 or later, or Terraform provider 5.29.0 or later.

Why it matters: Existing Workers VPC connections to private CAs that use weakened verification modes (verify_ca or disabled) can now use full certificate verification. Services requiring mTLS, including PostgreSQL databases with Amazon RDS CA bundles, become reachable where they were not before.

What to do: If you connect Workers VPC to private certificate authorities or databases requiring client certificates, upload your CA and client certificates with wrangler cert upload, reference them in VPC Service TLS settings, and set cert-verification-mode to verify_full to replace weakened verification.</description>
    </item>
    <item>
      <title>Cloudflare CLI 1.0.0-rc.1 moves build output to v1 directory, renames commands</title>
      <link>https://radar.trulysovereign.com/items/1b146c73-f687-487b-9370-ee26a5d296c9</link>
      <guid isPermaLink="false">1b146c73-f687-487b-9370-ee26a5d296c9</guid>
      <pubDate>Tue, 17 Nov 2026 20:59:55 GMT</pubDate>
      <category>migrate</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-17-release-candidate/">Cloudflare developer changelog</source>
      <description>Cloudflare CLI 1.0.0-rc.1 moves Build Output to .cloudflare/output/v1/ and renames several API commands without aliases (e.g., cf workers secrets bulk becomes cf workers secrets bulk-edit). The release requires @cloudflare/vite-plugin 2.0.0-rc.0+ for Vite projects and Wrangler 4.163.0+ for legacy projects. Command output behavior changed: prompts and status messages now write to stderr, with stdout reserved for command output only.

Why it matters: Scripts that parse cf command output will break if they read status messages from stdout instead of stderr. Automated workflows using renamed commands will fail without the new command names. Projects will fail to build if they use older versions of @cloudflare/vite-plugin or Wrangler than the new minimum requirements.

What to do: Update scripts that capture cf output to read status messages from stderr. Review the release notes for the complete list of renamed commands and update any automation or CI/CD pipelines that use them. Upgrade @cloudflare/vite-plugin to 2.0.0-rc.0+ or Wrangler to 4.163.0+ before updating cf.</description>
    </item>
    <item>
      <title>AI Search adds optional file summarization during indexing</title>
      <link>https://radar.trulysovereign.com/items/d2e51167-3de4-4ae9-88df-377aff9f6da1</link>
      <guid isPermaLink="false">d2e51167-3de4-4ae9-88df-377aff9f6da1</guid>
      <pubDate>Tue, 17 Nov 2026 20:59:43 GMT</pubDate>
      <category>budget</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-17-file-summaries/">Cloudflare developer changelog</source>
      <description>Cloudflare AI Search can now generate summaries, topics, and keywords for indexed text documents that produce at least three chunks. Enable by setting indexing_options.file_summary to true in the dashboard or API when creating or updating an instance. The feature is off by default, and changing the setting triggers a full reindex.

Why it matters: LLM summarization costs are billed separately as Workers AI usage, not covered by the AI Search ingestion allotment. Each new or modified file is summarized once, so costs scale with indexed content volume. Updating indexing_options replaces all options as a whole; omitted options revert to defaults.

What to do: If you update AI Search instances programmatically, include the current value of every indexing_options field in update calls to prevent unintended resets. If you enable file summaries, forecast the additional Workers AI costs based on your indexing volume.</description>
    </item>
    <item>
      <title>Cloudflare One Appliance HA sites can now configure separate WAN gateways per appliance</title>
      <link>https://radar.trulysovereign.com/items/58354c41-e1fd-4f23-8951-d8ec1e63a608</link>
      <guid isPermaLink="false">58354c41-e1fd-4f23-8951-d8ec1e63a608</guid>
      <pubDate>Tue, 17 Nov 2026 20:59:34 GMT</pubDate>
      <category>verify_only</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-11-17-appliance-ha-secondary-gateway/">Cloudflare developer changelog</source>
      <description>Cloudflare One Appliance high availability sites can now assign each appliance its own default gateway on a static WAN. The API accepts a new optional field static_addressing.secondary_gateway_address in PATCH requests to /accounts/{account_id}/magic/sites/{site_id}/wans/{wan_id}. The dashboard adds a Secondary gateway address field next to Primary gateway address when configuring static WANs on HA sites. If omitted or set to null, the secondary appliance uses the primary gateway address, matching previous behavior.

Why it matters: HA pairs can now connect each appliance to a separate ISP circuit with its own static IP and gateway, enabling true dual-circuit redundancy instead of sharing a single WAN gateway. Existing HA configurations continue working unchanged because the field defaults to the primary gateway when not specified.

What to do: If you operate Cloudflare One Appliance HA sites and want each appliance on a separate ISP circuit, add static_addressing.secondary_gateway_address to your WAN configuration API calls or populate the Secondary gateway address field in the dashboard. No action required if your HA pairs share a single WAN gateway.</description>
    </item>
    <item>
      <title>Cloudflare Email API uses 'address' field in REST, 'email' in Workers</title>
      <link>https://radar.trulysovereign.com/items/2b0475ae-1b06-43c5-bb73-49eaec2bd721</link>
      <guid isPermaLink="false">2b0475ae-1b06-43c5-bb73-49eaec2bd721</guid>
      <pubDate>Tue, 17 Nov 2026 20:59:23 GMT</pubDate>
      <category>migrate</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-05-28-named-email-recipients/">Cloudflare developer changelog</source>
      <description>The REST API for Cloudflare Email now uses 'address' as the field name for recipient email addresses, while the Workers binding continues to use 'email'. Both accept an optional 'name' field for display names on to, cc, bcc, replyTo, and from.

Why it matters: Code calling the REST API directly with the 'email' field will fail or send incorrectly. Workers binding calls are unaffected, creating a divergence between the two interfaces to the same service.

What to do: If you call the Cloudflare Email REST API, change recipient field names from 'email' to 'address'. Workers binding code requires no change.</description>
    </item>
    <item>
      <title>[CORRECTED] Cloudflare Workers Builds runtime defaults updated across Node.js, Python, Ruby, Go, Bun</title>
      <link>https://radar.trulysovereign.com/items/5f14d912-4b47-4c66-93f0-d81087bf4e04</link>
      <guid isPermaLink="false">5f14d912-4b47-4c66-93f0-d81087bf4e04</guid>
      <pubDate>Fri, 16 Oct 2026 16:42:44 GMT</pubDate>
      <category>pin_or_freeze</category>
      <category>corrected</category>
      <source url="https://developers.cloudflare.com/changelog/post/2026-10-14-workers-builds-default-tooling/">Cloudflare developer changelog</source>
      <description>[CORRECTED] Cloudflare Workers Builds changed default runtime versions effective 14 October 2026: Node.js 24.18.0→24.21.0, Python 3.13.3→3.14.8, Ruby 3.4.7→3.4.11, Go 1.24.3→1.27.2, Bun 1.2.15→1.4.3, Yarn 4.9.1→4.18.1, pip 25.1.1→26.2.1, pipx 1.7.1→1.17.2, Poetry 2.1.3→2.4.3, uv 0.7.13→0.12.10. Projects without explicit version pinning now build with these newer versions. Override environment variables are now documented: NODE_VERSION, PYTHON_VERSION, RUBY_VERSION, GO_VERSION, BUN_VERSION, YARN_VERSION.

Why it matters: Builds that relied on previous default versions may break if code or dependencies are incompatible with the newer runtimes. Python 3.13→3.14 and Go 1.24→1.27 are minor version jumps that can introduce breaking changes in language features or standard library behavior. Builds that worked yesterday may fail today without configuration changes.

What to do: Test your Workers builds immediately. If a build breaks or behaves differently, pin the previous version by setting the appropriate environment variable in your build configuration (e.g., PYTHON_VERSION=3.13.3, NODE_VERSION=24.18.0). Projects using .nvmrc, .node-version, .python-version, or .ruby-version files are unaffected and continue using their specified versions.

Original text (as first published, superseded):

WHAT CHANGED: Cloudflare Workers Builds changed default runtime versions effective 14 October 2026: Node.js 24.18.0→24.21.0, Python 3.13.3→3.14.8, Ruby 3.4.7→3.4.11, Go 1.24.3→1.27.2, Bun 1.2.15→1.4.3, Yarn 4.9.1→4.18.1, pip 25.1.1→26.2.1, Poetry 2.0.7→2.1.4, uv 0.7.13→0.12.10. Projects without explicit version pinning now build with these newer versions. Override environment variables are now documented: NODE_VERSION, PYTHON_VERSION, RUBY_VERSION, GO_VERSION, BUN_VERSION, YARN_VERSION.</description>
    </item>
  </channel>
</rss>
