Web Search API is now available in beta. Web Search API lets your AI agents and applications search the Internet and ground their responses in live information, instead of guessing URLs or relying on a model's training cutoff.
At launch, you can choose between three search providers: Ceramic.ai, Exa, and Linkup. All three support Zero Data Retention for requests made through Cloudflare, and all have committed to Cloudflare's verified bot crawling standards.
Web Search API runs through AI Gateway, so search requests appear in your gateway logs and are billed to your AI Gateway credits at each provider's list API price, with no additional markup. You can also bring your own provider API key.
Call Web Search API with the REST API:
curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/websearch/ \ --request POST \ --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \ --header "Content-Type: application/json" \ --data '{ "query": "What are some fun things to do in Salt Lake City as fall approaches?", "provider": "ceramic", "limit": 5, "options": { "gateway": { "id": "default" } } }'
Or from a Worker with the AI binding:
const response = await env.AI.websearch({ gatewayId: "default", query: "What are some fun things to do in Salt Lake City as fall approaches?", provider: "exa", limit: 5,});const results = await response.json();
The Agents SDK now provides first-class support for building agents using the Pi harness. You can build long-running agents using the combination of Pi 1.0 ↗︎, Pi Durable ↗︎, and the new PiHarness class that the Cloudflare Agents SDK provides, ensuring your agent's work is durably persisted, even if interrupted mid-turn.
Built with Earendil ↗︎, this integration is our first step toward first-class support for third-party agent harnesses on Cloudflare.
PiHarness is a new "Lifecycle capability" provided by the Cloudflare Agents SDK. Pi Durable provides the agent harness and the Lifecycle is responsible for keeping the agent running in the Durable Object. The Lifecycle is a core concept in the Agents SDK ensuring that long-running work can run in a Durable Object, surviving restarts, crashes, and network issues. We will share more on Lifecycle capabilities in the near future.
Install
npm i agents@latest @earendil-works/pi-durable @earendil-works/pi-ai
The agents/models/pi-ai entry point supports AI Gateway and Workers AI models, so you can get started with Cloudflare models right away or use your existing Pi AI provider.
Add tools with extensions
Both tools and system prompt sections are provided to the Pi Harness via extensions.
import { Type } from "@earendil-works/pi-ai";import { skills } from "agents/harness/pi";const WordCount = Type.Object({ text: Type.String() });const wordCount = { name: "word_count", description: "Count the words in a text.", parameters: WordCount, replay: "safe", async execute({ text }) { const words = text.split(/\s+/).filter(Boolean).length; return { content: [{ type: "text", text: String(words) }] }; },};// In the harness factory, before Harness.open():registry.install({ name: "editor", sections: [ { key: "preamble", render: () => "You are an editor.", tag: false }, ], tools: [wordCount],});registry.install(await skills(sources));
import { Type } from "@earendil-works/pi-ai";import type { ToolRegistration } from "@earendil-works/pi-durable";import { skills } from "agents/harness/pi";const WordCount = Type.Object({ text: Type.String() });const wordCount: ToolRegistration<typeof WordCount> = { name: "word_count", description: "Count the words in a text.", parameters: WordCount, replay: "safe", async execute({ text }) { const words = text.split(/\s+/).filter(Boolean).length; return { content: [{ type: "text", text: String(words) }] }; },};// In the harness factory, before Harness.open():registry.install({ name: "editor", sections: [ { key: "preamble", render: () => "You are an editor.", tag: false }, ], tools: [wordCount],});registry.install(await skills(sources));
For more information on creating and configuring extensions, refer to Extensions.
Jurisdictions for Workers KV namespaces are now generally available. When you create a namespace, you can set a jurisdiction to make sure the namespace's data is only durably stored within that region. Jurisdictions can help you comply with data localization regulations such as GDPR or FedRAMP. Supported jurisdictions are eu, us, and fedramp.
A jurisdiction can only be set when a namespace is created, using the Cloudflare dashboard, Wrangler, the cf CLI, or the REST API, and cannot be added or changed afterwards.
Workers can still access a namespace restricted to a jurisdiction from anywhere in the world, and KV data can be cached outside the jurisdiction on Cloudflare's network. The jurisdiction only controls where the namespace's data is durably stored.
You can now restrict who can access a Quick Tunnel. Use the new --allowed-mail flag in cloudflared to require visitors to authenticate with a one-time PIN sent to their email before they reach your local service.
Previously, anyone with a trycloudflare.com URL could access the service behind it. Protected Quick Tunnels let you share a local development server, webhook receiver, or demo with specific people without creating a Cloudflare account or configuring a domain.
You can allow:
A single email address: --allowed-mail alice@example.com
Multiple email addresses, by repeating the flag or using a comma-separated list: --allowed-mail 'alice@example.com,bob@example.com'
Every address on a domain: --allowed-mail '*@example.com'
Visitors do not need a Cloudflare account. Access ends for everyone when you stop the cloudflared process.
Clef is a decision model, in the same family as Typesafe's Jev ↗︎. Instead of generating text, it reads an input state and a set of typed questions, then returns a probability for every allowed answer. Your agent gets a structured decision it can act on immediately, for example: route the ticket, block the request, or escalate to a human. There is no free-form output to parse and no reasoning tokens to wait for.
Both models are hosted on Workers AI as Clef and Clef-flash. We are also open-sourcing the weights under the Apache 2.0 license on Hugging Face: Clef ↗︎ and Clef-flash ↗︎. Read the launch blog post ↗︎ for the full story, including how we trained them.
Clef is designed to be fast so decisions come back in milliseconds. Across our 43 benchmark runs, we achieved speeds where Clef is 2.5x faster than Jev at the median, and Clef-flash 13x faster.
Latency
Clef
Clef-flash
Jev
Median
209.3 ms
38.8 ms
524.1 ms
p95
238.6 ms
122.4 ms
536.0 ms
Hosting on Workers AI adds to that speed. Requests run on GPUs across Cloudflare's network, running close to your users, so the network round trip stays short. You can put Clef directly in the request path of your agent, then hand off to an LLM on Workers AI to take action.
Leading the benchmarks
Across 10 decision benchmarks, a Clef model scores highest on 7, ahead of Jev and other open decision models. A few highlights:
Benchmark
Clef
Clef-flash
Jev
BFCL (case exact)
98.47
98.76
95.75
BANKING77 (macro-F1)
94.20
90.93
79.74
CLINC150+OOS (macro-F1)
97.43
66.77
89.27
Home appliances (case exact)
82.95
97.73
52.27
On Typesafe's own workflow evals, Clef beats Jev in 3 of 4 areas: invoice processing, customer service, and security incidents. The full results are on the Hugging Face model card ↗︎.
Clef follows the System One API, so you can switch an existing Jev integration to Clef by changing the endpoint and model. Ask up to 64 questions per request, in three types:
noul: A yes/no question. Returns the probability that the answer is yes.
choice: Pick one option from a set you define. Returns the chosen option, a probability per option, and a confidence value.
score: Rate against an ordered rubric. Returns a probability-weighted score and a probability per level.
const response = await env.AI.run("@cf/cloudflare/clef", { model: "clef", state: "Checkout has been failing for every customer for the last hour.", questions: { urgent: { type: "noul", instructions: "Is this support request urgent?", }, team: { type: "choice", instructions: "Which team should handle this request?", criteria: { billing: "Payments, invoices, and refunds", technical: "Outages, errors, and configuration", sales: "Plans and upgrades", }, }, },});// response.answers.urgent.noul -> probability the request is urgent// response.answers.team.choice -> highest-probability team
const response = await env.AI.run("@cf/cloudflare/clef", { model: "clef", state: "Checkout has been failing for every customer for the last hour.", questions: { urgent: { type: "noul", instructions: "Is this support request urgent?", }, team: { type: "choice", instructions: "Which team should handle this request?", criteria: { billing: "Payments, invoices, and refunds", technical: "Outages, errors, and configuration", sales: "Plans and upgrades", }, }, },});// response.answers.urgent.noul -> probability the request is urgent// response.answers.team.choice -> highest-probability team
What you can build with decision models
Support triage: Decide whether a ticket is urgent and which team owns it, then route it without a human in the loop.
Threat intelligence: Classify a website by category. Paired with Browser Run, Clef fetched, rendered, and classified a domain in 2.2 seconds, compared to 4.7 seconds for gpt-oss-120b in the same workflow.
Trust and safety: Score user submissions against your own policy rubric and act on the probability.
Agent guardrails: Let an agent check "should I take this action?" in tens of milliseconds before calling a tool.
Visual classification: Pass up to four images alongside the state. Unlike text-only decision models, Clef has a vision encoder.
Get started
Use Clef through the Workers AI binding (env.AI.run()) or the REST API at /ai/run. You can also use AI Gateway with these endpoints.
@cloudflare/workers-oauth-provider ↗︎ is now v1, with a new split API. One Worker acts as the authorization server: it signs users in and issues tokens. Your MCP server acts as the resource server, and can run in another Worker. It validates each token with the authorization server over a Service Binding, without crossing the public Internet.
A migration skill ↗︎ ships in the npm package so a coding agent can perform the upgrade.
The split API
import { OAuthAuthorizationServer, OAuthResourceServer, insufficientScope,} from "@cloudflare/workers-oauth-provider";import { WorkerEntrypoint } from "cloudflare:workers";// auth-server Worker: signs users in and issues tokens for both MCP servers.const authorizationServer = new OAuthAuthorizationServer({ issuer: "https://auth.example.com", resources: [ "https://calendar.example.com/mcp", "https://drive.example.com/mcp", ], scopesSupported: ["calendar:read", "calendar:write", "offline_access"], clientIdMetadataDocumentEnabled: true,});export class AuthServer extends WorkerEntrypoint { fetch(request) { if (new URL(request.url).pathname === "/authorize") { return showConsent(request, this.env); } return authorizationServer.fetch(request, this.env, this.ctx); } validateToken(resource, token) { return authorizationServer.validateToken(resource, token, this.env); }}// calendar MCP Worker: checks tokens with AuthServer over a Service Binding.export const calendar = new OAuthResourceServer({ resourceMetadata: { resource: "https://calendar.example.com/mcp", authorization_servers: ["https://auth.example.com"], }, requiredScopes: ["calendar:read"], validateToken: (env) => env.AUTH_SERVER.validateToken, handler: { fetch(request, env, ctx) { if ( request.method === "POST" && !ctx.auth.scope.includes("calendar:write") ) { return insufficientScope(ctx.auth, ["calendar:read", "calendar:write"]); } return handleMcp(request, ctx.props); }, },});
import { OAuthAuthorizationServer, OAuthResourceServer, insufficientScope,} from "@cloudflare/workers-oauth-provider";import { WorkerEntrypoint } from "cloudflare:workers";// auth-server Worker: signs users in and issues tokens for both MCP servers.const authorizationServer = new OAuthAuthorizationServer<Env>({ issuer: "https://auth.example.com", resources: [ "https://calendar.example.com/mcp", "https://drive.example.com/mcp", ], scopesSupported: ["calendar:read", "calendar:write", "offline_access"], clientIdMetadataDocumentEnabled: true,});export class AuthServer extends WorkerEntrypoint<Env> { fetch(request: Request) { if (new URL(request.url).pathname === "/authorize") { return showConsent(request, this.env); } return authorizationServer.fetch(request, this.env, this.ctx); } validateToken(resource: string, token: string) { return authorizationServer.validateToken(resource, token, this.env); }}// calendar MCP Worker: checks tokens with AuthServer over a Service Binding.export const calendar = new OAuthResourceServer<Env, AuthProps>({ resourceMetadata: { resource: "https://calendar.example.com/mcp", authorization_servers: ["https://auth.example.com"], }, requiredScopes: ["calendar:read"], validateToken: (env) => env.AUTH_SERVER.validateToken, handler: { fetch(request, env, ctx) { if ( request.method === "POST" && !ctx.auth.scope.includes("calendar:write") ) { return insufficientScope(ctx.auth, ["calendar:read", "calendar:write"]); } return handleMcp(request, ctx.props); }, },});
In the example, env.AUTH_SERVER.validateToken is that Service Binding call. The calendar Worker needs no KV namespace of its own.
{ "name": "calendar-mcp", "main": "src/index.ts", // Set this to today's date "compatibility_date": "2026-10-20", "services": [ { "binding": "AUTH_SERVER", "service": "auth-server", "entrypoint": "AuthServer", }, ],}
name = "calendar-mcp"main = "src/index.ts"# Set this to today's datecompatibility_date = "2026-10-20"[[services]]binding = "AUTH_SERVER"service = "auth-server"entrypoint = "AuthServer"
OAuthResourceServer publishes the RFC 9728 ↗︎ protected resource metadata that MCP clients use to find your authorization server. It answers requests without a token with a 401 challenge that points to that metadata. It also rejects tokens issued for any other resource.
You can still use OAuthProvider as both the authorization server and the MCP server. For most 0.x deployments, the only required change is to add resourceMetadata: { resource }.
AI Search is now generally available. Usage-based billing begins on November 1, 2026, with included monthly ingestion, storage, semantic query, and full-text query usage. Cloudflare will send a reminder email the week before billing begins.
New AI Search instances use hybrid search by default. Hybrid search combines semantic vector retrieval with full-text matching. You can choose a different index method when you create an instance.
Workers AI embedding and reranking calls made by AI Search are included in AI Search pricing. These calls no longer appear on your Workers AI bill or in your AI Gateway logs. Generation, query rewriting, and external providers continue to use your account and gateway.
AI Search supports the @cf/qwen/qwen3-vl-embedding-2b and google-ai-studio/gemini-embedding-2 multimodal embedding models. Search and chat requests can include images through the REST API and public endpoint.
Refer to Supported models for the full list of embedding models.
OCR availability and increased file limits
Optical character recognition (OCR) is available on every account for scanned PDFs. Plain-text or code files and PDFs with OCR enabled can be up to 10 MiB. PDFs without OCR and other supported formats remain limited to 4 MiB.
When you create an AI Search instance, the type field is optional. AI Search infers a website source from an HTTP or HTTPS URL, or an R2 source from an existing bucket name.
Artifacts, Cloudflare's versioned file system that speaks Git, is now in open beta. Artifacts is built for scale, so you can create a repository per project, user, session, or task.
With Artifacts, you can:
Deploy repositories to Workers — Connect an Artifacts repository through Workers Builds. Pushes to the production branch deploy the updated Worker, while other branches create or update Worker Previews.
Programmatically manage repositories — Use an Artifacts binding from a Worker to create or fork repos, inspect files and commits, read files by path, and issue repo-scoped Git tokens.
React to repository changes — Subscribe to events when a repository is created, imported, forked, deleted, pushed to, cloned, or fetched.
Control where repository data is stored — Choose to store and process your data in the US or EU.
Monitor repository usage — View total operations, pulls, pushes, errors, and error rates in the Cloudflare dashboard or via API for analytics.
Artifacts is available for customers on the Workers Paid plan. Cloudflare will begin billing for Artifacts on October 14, 2026.
Build the next GitHub on Cloudflare
We are hosting a competition to see who can build the next GitHub on Cloudflare using Workers and Artifacts.
Apply today ↗︎ — submissions are open until October 14, 2026.
The first-place team will receive $25,000 in Cloudflare credits. The top three teams will be flown to San Francisco to present what they built at Cloudflare Connect.
The higher per-stream limit gives high-volume application events, telemetry, and logs more room to grow without splitting ingestion across streams solely to stay within the previous limit.
For the full list of stream, sink, and pipeline limits, refer to Basin Pipelines limits.
Basin, formerly the Cloudflare Data Platform, is now generally available. Basin brings an end-to-end analytics platform to the Developer Platform, enabling you to collect data from a variety of sources, such as apps, infrastructure, devices, and other Cloudflare services, then query it to answer analytical questions.
Basin Pipelines
Basin Pipelines, formerly Cloudflare Pipelines, ingests events from Workers, HTTP endpoints, and Cloudflare Logpush. It transforms events with SQL and ingests them into Iceberg tables or files on R2. With Basin Pipelines you can:
Ingest application and device events through HTTP endpoints or Workers bindings.
Filter and reshape Cloudflare logs before storing them as Iceberg tables, Parquet, or JSON.
Catch schema mismatches with typed bindings and investigate dropped events in the dashboard.
Basin Catalog
Basin Catalog, formerly R2 Data Catalog, manages and automatically maintains Apache Iceberg tables ↗︎ to keep them fast, cost-efficient, and accessible to any compatible query engine. With Basin Catalog you can:
Connect DuckDB, Spark, Snowflake, or PyIceberg to the same tables.
Maintain growing tables with compaction, snapshot expiration, and manifest optimization.
Share analytical data across tools and clouds without paying egress fees.
Basin SQL
Basin SQL, formerly R2 SQL, is a serverless, distributed SQL engine for querying large Apache Iceberg tables in Basin Catalog without managing or scaling compute. With Basin SQL you can:
Summarize and rank data with standard and approximate aggregates, grouping sets, and window functions.
Combine and inspect datasets with joins, subqueries, common table expressions, set operations, schema discovery, and EXPLAIN.
Transform strings, timestamps, JSON, and complex values with more than 190 functions.
Get started
To get started with creating an end-to-end data pipeline, run:
Durable Objects remain active while handling a request from a connected client. This change applies when no client is connected, such as when an agent continues a submitted job after its client disconnects.
This behavior is the default for Workers with a compatibility date of 2026-10-01 or later. To use it with an earlier date, add the durable_object_io_tasks_prevent_eviction compatibility flag. To opt out, add the durable_object_io_tasks_do_not_prevent_eviction flag.
Pending service binding requests now keep Durable Objects running while they wait for a response. Pending calls to another Durable Object through remote procedure call (RPC) or fetch(), as well as this.ctx.container.monitor(), now also keep the Durable Object running.
Promises passed to this.ctx.waitUntil() and pending setTimeout() and setInterval() timers also receive this protection.
Previously, Cloudflare could shut down an idle Durable Object while one of these operations remained pending without a connected client. This could stop unfinished work.
This change helps you run long-running tasks such as agents. An agent can call tools through service bindings, coordinate with other Durable Objects, or wait for a container process without relying on the original client to remain connected.
Outbound fetch() requests to external services, TCP sockets, and outbound WebSockets already keep Durable Objects running.
Each pending operation prevents idle shutdown for up to 15 minutes. Starting another one later can extend the Durable Object's time in memory. The limit applies to each operation, not to the total time in memory.
Duration charges continue while an operation prevents eviction.
The Workers Web Crypto API now supports ML-KEM-768, ML-KEM-1024, ML-DSA-44, ML-DSA-65, and ML-DSA-87. ML-KEM establishes shared secrets, while ML-DSA signs and verifies data.
The opt-in API also adds key encapsulation and decapsulation methods, getPublicKey(), SubtleCrypto.supports(), and JSON Web Keys (JWKs) with the AKP key type.
Turn on the webcrypto_modern_algorithms compatibility flag to use these features:
Workers implements a subset of the evolving Modern Algorithms in the Web Cryptography API ↗︎ draft. ML-KEM-512 and the draft's other algorithms are not supported. The API may change as the draft evolves.
Containers now support the durable_object scheduling policy in public beta. This policy lets a Durable Object select the image and instance size for a Container at runtime instead of using one centrally managed configuration for the application.
To use custom images, configure the policy and one or more named images in Wrangler:
Wrangler prepares each image and exposes its immutable reference through ctx.container.images. Supply that reference and an instance size when you start the Container:
The durable_object policy also supports the new cloudflare/debian-trixie Cloudflare-managed image, which includes Node.js 24.20.0 on Debian Trixie slim. Start it directly without configuring a named image.
Durable Object-managed Container instances have independent lifecycles and do not participate in application-wide image rollouts.
For configuration, runtime sizing, snapshots, and update behavior, refer to Scheduling Policies.
Containers now support snapshot APIs in public beta for saving and restoring point-in-time filesystem state. Create a snapshot first, then pass it back to start() to restore files after container sleep, restart, or handoff to another Durable Object.
Use snapshotContainer() through the Durable Object Container API to capture the full container filesystem. Create a snapshot from a running Container, store its handle, and pass that handle to start() when you restore it later:
src/index.jsjs
import { DurableObject } from "cloudflare:workers";export class MyDurableObject extends DurableObject { async saveSnapshot() { // Create a snapshot from the running Container. const containerSnapshot = await this.ctx.container.snapshotContainer({}); await this.ctx.storage.put("containerSnapshot", containerSnapshot); } async restoreSnapshot() { // Restore the saved snapshot later. const containerSnapshot = await this.ctx.storage.get("containerSnapshot"); if (!containerSnapshot) { return; } this.ctx.container.start({ containerSnapshot, enableInternet: false }); }}
src/index.tsts
import { DurableObject } from "cloudflare:workers";export class MyDurableObject extends DurableObject { async saveSnapshot() { // Create a snapshot from the running Container. const containerSnapshot = await this.ctx.container.snapshotContainer({}); await this.ctx.storage.put("containerSnapshot", containerSnapshot); } async restoreSnapshot() { // Restore the saved snapshot later. const containerSnapshot = await this.ctx.storage.get<ContainerSnapshot>("containerSnapshot"); if (!containerSnapshot) { return; } this.ctx.container.start({ containerSnapshot, enableInternet: false }); }}
Snapshots are only supported for Container applications that use the durable_object scheduling policy. Snapshots are immutable, so create a new snapshot to persist filesystem changes made after a restore.
Sandbox SDK 1.0 is available. Your own Durable Object class now controls each sandbox container directly, through the Durable Object container API on this.ctx.container.
With 1.0, your class can:
Choose the image and instance size each time it starts a sandbox. One class can run sandboxes on different images, and a deploy does not restart sandboxes that are running.
Save the files of a sandbox as a snapshot, in public beta, and start the same sandbox or a new one from it.
Decide when each sandbox stops, for example when a task finishes, when its user goes idle, or after it saves a snapshot.
Run commands with streamed input and output, send them signals, and open terminals.
Serve previews from ports in the sandbox, with your own hostnames and authentication.
Handle outbound requests for each hostname in Worker code, so credentials and bindings stay in your Worker.
Expose only the methods that you want callers to use.
Run an agent in the same Durable Object, and give the model a tool that runs commands in the sandbox.
src/index.jsjs
import { Files } from "@cloudflare/sandbox";import { DurableObject } from "cloudflare:workers";export class MySandbox extends DurableObject { container; files; constructor(ctx, env) { super(ctx, env); if (!ctx.container) { throw new Error("No container is configured"); } this.container = ctx.container; this.files = new Files(ctx.container); } async run(script) { if (!this.container.running) { // Your code chooses the image, size, and network access. this.container.start({ image: this.container.images.sandbox, instance: "lite", enableInternet: false, }); } await this.files.writeFile("/tmp/task.sh", script); const proc = await this.container.exec(["sh", "/tmp/task.sh"]); return proc.output(); }}
src/index.tsts
import { Files } from "@cloudflare/sandbox";import { DurableObject } from "cloudflare:workers";export class MySandbox extends DurableObject<Env> { private readonly container: Container; private readonly files: Files; constructor(ctx: DurableObjectState, env: Env) { super(ctx, env); if (!ctx.container) { throw new Error("No container is configured"); } this.container = ctx.container; this.files = new Files(ctx.container); } async run(script: string) { if (!this.container.running) { // Your code chooses the image, size, and network access. this.container.start({ image: this.container.images.sandbox, instance: "lite", enableInternet: false, }); } await this.files.writeFile("/tmp/task.sh", script); const proc = await this.container.exec(["sh", "/tmp/task.sh"]); return proc.output(); }}
Files streams files in and out of the running sandbox.
S3Mount mounts an S3-compatible bucket, such as R2. Your Worker signs each storage request, so the credentials stay out of the sandbox.
DirectoryBackup saves a directory to R2 and restores it into any sandbox, including one on a newer image.
If you use Sandbox SDK 0.x
Your 0.x applications keep running, and @cloudflare/sandbox 0.x stays on npm. Sandbox SDK 0.x receives bug and security fixes until 2026-12-31, and its documentation stays at Sandbox SDK 0.x.
When you are ready, Migrate from Sandbox SDK 0.x shows the 1.0 code for each 0.x feature, including preview URLs, tunnels, background processes, terminals, backups, and the code interpreter. The guide keeps the preview URLs and named tunnels that your 0.x application created working. You can move every sandbox in one deploy, or run a 1.0 class next to your 0.x class and move sandboxes one at a time. The deploy that moves an existing class to the new policy is one-way, so the guide shows how to rehearse it first.
The Sandboxes documentation also covers Dynamic Workers, for untrusted code in JavaScript, Python, or WebAssembly.
AI Gateway now supports Machine Payments in beta. With Machine Payments, clients can use the x402 protocol to pay for eligible inference requests directly from a stablecoin wallet instead of maintaining a prepaid credit balance.
Machine Payments is available for the /ai/run endpoint with select open models. To request x402 payment, authenticate with a Cloudflare API token and include the Cloudflare-specific Payment-Method: x402 header:
An x402-compatible client handles the payment challenge, signs an authorization from the client's wallet, and retries the request. Machine Payments currently requires customers to be based in the United States and have a credit card on file.
For prerequisites, eligible models, and transaction details, refer to Machine Payments (x402).
The adapter connects live calls to any server that accepts WebSockets, such as a Durable Object. It delivers uncompressed pulse-code modulation (PCM) audio and JPEG video frames. Your backend can process this media without implementing a WebRTC client.
What you can build
Transcribe or record call audio, or let a voice agent hear and respond to participants. The AI audio example uses Durable Objects and Workers AI for transcription and speech generation. Separate adapters receive PCM audio and send generated speech back to participants.
Analyze images or build previews from live video. The WebRTC-to-JPEG example ↗︎ sends a browser's camera stream to a Durable Object as JPEG frames at one frame per second by default.
What changes for existing integrations
Existing adapter creation requests and media formats stay the same.
For WebRTC-to-WebSocket streaming, the SFU now automatically retries the same endpoint for up to 15 seconds instead of 5, with no additional API setting.
The close API is now idempotent and returns success even if the adapter has already closed:
AI Gateway User Insights now gives you more context about the traffic flowing through your gateway. It shows what users and agents are doing with AI, and where a selected model may be more capable than a task requires.
On the analysis side, User Insights groups conversations by task, tracks conversation turns, and helps you compare model fit with cost and latency.
The Potential Savings view highlights requests that may work with faster or less expensive models without compromising output quality. These are the same signals that Cloudflare's Auto Router uses to select a model based on task and cost.
These new insights are available to all AI Gateway customers at no additional cost. For more information, refer to User Insights.
Browser Run sessions now accept multiple concurrent connections. Before, a session accepted only one connection at a time, and other Workers had to wait until that connection closed. Now multiple Workers can connect to the same browser at the same time.
Each puppeteer.connect() call opens its own Chrome DevTools Protocol (CDP) connection. Create a separate browser context for each request to keep its pages, cookies, and storage apart from other clients.
Containers custom instance types no longer limit disk based on memory. Previously, a custom instance type could have a maximum of 2 GB of disk for each 1 GiB of memory. You can now allocate up to the 20 GB disk maximum to any custom instance type.
Use this to run workloads that need more disk than memory, such as workloads with large container images, datasets, or build caches. The maximum image size is the same as the instance disk space, so more disk also lets you deploy larger images.
For example, a custom instance type with 1 vCPU and 3 GiB of memory was previously limited to 6 GB of disk. It can now use 20 GB:
The other custom instance type constraints do not change, including the minimum of 3 GiB of memory per vCPU. For the full list, refer to Custom Instance Types.
You can now tell a person on a laptop apart from a Mesh node or an AI agent running on Workers, without matching on connector email addresses or Mesh IP ranges — and see exactly which Cloudflare Tunnel and cloudflared replica received each session.
Mesh — Traffic sent from or delivered to a Cloudflare Mesh node. Previously, Mesh nodes were logged the same way as devices running the Cloudflare One Client, because Mesh nodes run the client in headless mode.
Workers VPC — Traffic sent by a Worker through a Workers VPC binding. Previously, Workers VPC sessions were not recorded in Network Session Logs.
Gateway network logs
To view these values in the dashboard, go to Zero Trust > Insights & Logs > Logs > Network logs, select Columns, and turn on Traffic Source and Traffic Destination. Both values also appear under Network query details when you open a log entry.
Network Session Logs
The zero_trust_network_sessions dataset, available through Logpush, includes the following fields:
Field
Description
OnrampType
How the session entered Cloudflare One. Values: CF1_CLIENT, MESH, WORKERS_VPC, MAGIC, OTHER.
Offramp
Where the session was routed. Sessions routed to a Mesh node report MESH.
SourceName
Name of the Worker that started the session. Only populated for Workers VPC sessions.
SourceID
Stable identifier of the Worker that started the session. Only populated for Workers VPC sessions.
DestinationReplicaID
The replica that served the session, such as a specific replica of a Mesh node or a cloudflared replica of a Cloudflare Tunnel.
For example, OnrampType = 'WORKERS_VPC' AND Offramp = 'MESH' returns every session where a Worker reached a service behind a Mesh node, and SourceName tells you which Worker it was.
See which tunnel and replica received a session
With DestinationReplicaID, you can now confirm which Cloudflare Tunnel and which cloudflared replica received traffic for a specific session. Combine it with the existing DestinationTunnelID field to trace a session to an exact tunnel replica — or Mesh node replica — when you run multiple replicas for high availability. The replica ID matches the Connector ID shown in the dashboard, so you can stream that replica's logs with cloudflared tail --connector-id.
Workers Cache now supports invalidate(), the soft counterpart of purge(). purge() deletes matching cached responses, so the next request is a cache miss. invalidate() keeps them but marks them stale, so the cache revalidates them with your Worker instead.
To revalidate a response, the cache sends your Worker a conditional request built from the validators stored with it — for example, If-None-Match carrying the cached ETag. If your Worker answers 304 Not Modified, the cache keeps the stored body. If your Worker answers with a full 200 response, that response replaces the cached one.
invalidate() accepts the same options as purge(): tags, pathPrefixes, or purgeEverything. It follows the same per-entrypoint scoping and resolves to the same result object. Call it as ctx.cache.invalidate(), or import cache from cloudflare:workers and call cache.invalidate().
Use invalidate() when one call covers many cached responses but only some of them changed. Your Worker needs to emit ETag or Last-Modified and answer matching conditional requests with 304. Each unchanged response then costs a validator check instead of a full regeneration:
src/index.jsjs
export default { async fetch(request, env, ctx) { if (request.method === "POST") { // Write the updated catalog, then mark every cached product page stale. await syncCatalog(env, await request.json()); await ctx.cache.invalidate({ tags: ["products"] }); return new Response("Synced"); } const product = await getProduct(env, request); const etag = `"${product.revision}"`; const headers = { "Cache-Control": "public, max-age=86400", "Cache-Tag": "products", ETag: etag, }; // The product has not changed since it was cached. Answer 304, and the // cache keeps the body it already has. if (request.headers.get("If-None-Match") === etag) { return new Response(null, { status: 304, headers }); } return new Response(renderProductPage(product), { headers }); },};
src/index.tsts
export default { async fetch(request, env, ctx): Promise<Response> { if (request.method === "POST") { // Write the updated catalog, then mark every cached product page stale. await syncCatalog(env, await request.json()); await ctx.cache.invalidate({ tags: ["products"] }); return new Response("Synced"); } const product = await getProduct(env, request); const etag = `"${product.revision}"`; const headers = { "Cache-Control": "public, max-age=86400", "Cache-Tag": "products", ETag: etag, }; // The product has not changed since it was cached. Answer 304, and the // cache keeps the body it already has. if (request.headers.get("If-None-Match") === etag) { return new Response(null, { status: 304, headers }); } return new Response(renderProductPage(product), { headers }); },} satisfies ExportedHandler<Env>;
