Recent releases from organizations and products in the Infrastructure category.
OpenAI's GPT-6 Astra model is now available through Netlify's AI Gateway and Agent Runners with zero configuration, working directly with the OpenAI SDK in Netlify Functions. It's also supported across Background Functions, Scheduled Functions, and Edge Functions, with automatic access to caching, rate limiting, and authentication.
Agent Runners now scope ambitious prompts for new projects, building a partial working project and suggesting next steps within the available credit budget instead of stalling out mid-run.
Wrangler no longer checks compressed bundle size — the 3 MB (Free) and 10 MB (Paid) limits are removed. Cloudflare now enforces a single 64 MiB uncompressed limit across all plans. Use wrangler deploy --outdir bundled/ --dry-run to check your bundle's uncompressed size before deploying.
Python Workers now support web frameworks following the WSGI or ASGI specification, including Django, Flask, FastAPI, and Starlette. Use wsgi or asgi from the workers module with entrypoint() for a default setup, or drop to WorkerEntrypoint for more control.
Cursor Cloud Agents can now run on Cloudflare via self-hosted machines, with each session in an isolated environment backed by Cloudflare Containers. Cloudflare executes commands, file edits, and repository operations on infrastructure you control, while Cursor hosts the agent loop and inference.
Cloudflare's Images binding now supports rendering text into images, filtering hosted images by metadata, generating signed URLs for private images, and accepting user uploads via Direct Creator Uploads. The .response() method also now accepts a headers option for simplified response building.
The Routes page now supports creating multiple Cloudflare Tunnel and Cloudflare Mesh routes in a single action, including comma-separated destinations, queued routes, and retrying only failed routes. Bulk creation for Cloudflare WAN static routes is also available.
define custom applications for breakout and prioritized traffic directly from the Appliance dashboard's Traffic Steering panel, with new Source subnets matching, without calling the API. Edit or delete existing custom applications in the same panel.
DHCP options can now be configured directly from the dashboard when the Cloudflare One Appliance acts as the LAN's DHCP server, complementing the existing API and Terraform workflow. Options include common PXE/iPXE boot, VoIP provisioning, and vendor-specific settings, or custom option codes.
You can now create multiple Cloudflare Tunnel and Cloudflare Mesh routes from the Routes page in a single action, instead of submitting one at a time. When creating a route you can add multiple destinations as a comma-separated list, queue up additional routes before creating them all together, and retry only the routes that failed in a batch since successful ones are removed from the form automatically.
Google's Gemini 3.8 Flash model is now available in Netlify's AI Gateway and Agent Runners with zero configuration. Use the Google GenAI SDK directly in Netlify Functions without managing API keys, with automatic access to caching, rate limiting, and authentication.
Anthropic's Claude Fable 5.1 model is now available through Netlify's AI Gateway and Agent Runners with zero configuration. Use the Anthropic SDK directly in Netlify Functions without managing API keys or authentication; AI Gateway handles caching, rate limiting, and auth automatically.
Adds a new threat detection to the Cloudflare Managed Ruleset, improving coverage for SQL injection patterns combining WHERE comparisons with WITH clauses. The new rule, SQLi - WHERE Comparison With WITH Clause, transitions from Log to Block action.
AI Gateway monthly usage invoices now show a single total cost per model instead of separate input and output token line items, using standardized provider/model identifiers. This does not apply to invoices for AI Gateway credit purchases.
Beginning September 1, 2026, D1 queries on the Workers Free plan fail when an account exceeds daily row read or write limits, returning errors until midnight UTC. Stored data is not affected; email alerts notify on limit reach. Optimization steps include adding indexes and upgrading to a paid plan for higher limits.
The /crawl endpoint now respects the use directive of the Content Signals standard, letting site owners express the maximum level at which their content may be used. A new contentUse parameter (values reference and full, default full) declares your intended level; if a target site's robots.txt sets a more restrictive use level, the crawl is rejected with a 400 error.
AI Search now supports the @cf/zai-org/glm-5.3-flash model for text generation, featuring a 1,048,576-token context window on Workers AI. Configuration details are in the Supported models documentation.
