
TL;DR
Cloudflare shipped wrangler deploy --temporary on June 19, 2026. AI agents can now deploy Workers, D1 databases, and KV stores without browser auth flows. Here is how it works.
| Source | Link |
|---|---|
| Cloudflare Blog Announcement | blog.cloudflare.com/temporary-accounts |
| Cloudflare Changelog | developers.cloudflare.com/changelog |
| Wrangler CLI Documentation | developers.cloudflare.com/workers/wrangler |
| Workers Documentation | developers.cloudflare.com/workers |
Last updated: June 20, 2026
AI agents hit a wall when they need to deploy code. The problem is not the deployment itself - it is the account creation flow that precedes it. OAuth popups, dashboard click-throughs, API token copy-paste rituals, multi-factor authentication challenges - all designed for humans with browsers and patience.
Cloudflare addressed this on June 19, 2026 with temporary accounts. An agent running Wrangler 4.102.0 or later can now deploy a Worker, a D1 database, KV storage, Durable Objects, or any supported Cloudflare resource without creating an account first. The deployment stays live for 60 minutes. During that window, a human can claim the account and make it permanent.
The mechanism is a single CLI flag:
wrangler deploy --temporary
When an agent runs this command without credentials, Wrangler provisions a temporary Cloudflare account in the background, issues an API token to itself, deploys the Worker, and returns two URLs: the live deployment URL and a claim URL.
The agent does not need to navigate a browser, handle OAuth callbacks, or parse emails. It gets infrastructure.
Temporary deployments self-destruct after 60 minutes of inactivity. That is enough time for an agent to:
Each redeploy resets the 60-minute timer. An agent actively iterating on a deployment can keep it alive indefinitely - the clock only runs when nothing is happening.
When a human clicks the claim URL, they authenticate with Cloudflare (signing up or logging in), and the temporary account converts to a permanent one. All resources - Workers, databases, KV namespaces, Durable Objects - transfer with it.
If nobody claims the account within the window and no redeploy happens, Cloudflare deletes everything.
The temporary accounts feature supports a significant slice of the Cloudflare developer platform:
| Product | Supported |
|---|---|
| Workers | Yes |
| Workers Static Assets | Yes |
| Workers KV | Yes |
| D1 (SQLite database) | Yes |
| Durable Objects | Yes |
| Hyperdrive | Yes |
| Queues | Yes |
| SSL/TLS certificates | Yes |
This covers most common deployment patterns. An agent can spin up an API with a database backend, deploy a static site, or launch a real-time application with Durable Objects - all without a pre-existing account.
Get the weekly deep dive
Tutorials on Claude Code, AI agents, and dev tools - delivered free every week.
From the archive
Jun 20, 2026 • 9 min read
Jun 20, 2026 • 6 min read
Jun 20, 2026 • 6 min read
Jun 20, 2026 • 11 min read
You need Wrangler 4.102.0 or later. Check your version:
wrangler --version
If you are behind, update:
npm install -g wrangler@latestOne gotcha: if you have cached credentials from a previous login, they may interfere with temporary account creation. Run wrangler logout before testing the feature to ensure a clean state.
The pattern Cloudflare enabled here is "deploy first, claim later." It inverts the traditional flow where account creation is a prerequisite for any action.
For agent developers, this opens several use cases:
Prototype deployment agents. An agent can take a description like "build me an API that converts markdown to PDF" and ship a working endpoint without any credential handoff. The human reviews the live deployment before claiming it.
CI/CD without secrets. Ephemeral preview deployments in CI pipelines can use temporary accounts instead of storing Cloudflare API tokens in secrets. The deployment lives long enough for automated tests, then either gets claimed or expires.
Sandbox environments. Agents testing infrastructure code can deploy to throwaway Cloudflare accounts, validate behavior, and clean up automatically after 60 minutes.
Demo flows. A conversational agent can deploy a working example during a support interaction. The user sees it live, decides if they want it, and claims or abandons it.
Cloudflare built in a helpful fallback. When an agent tries to run wrangler deploy without credentials, Wrangler outputs a message suggesting the --temporary flag. An agent parsing CLI output can discover the feature and adapt without explicit instructions.
This matters because it means agents designed for credentialed deployments can gracefully degrade to temporary deployments when credentials are unavailable. No special-case code required - just read the error message.
Here is a minimal agent loop that deploys a Worker and returns both URLs:
import { execSync } from 'child_process';
async function deployTemporary(projectPath: string) {
// Ensure no cached credentials
execSync('wrangler logout', { cwd: projectPath, stdio: 'ignore' });
// Deploy with temporary flag
const output = execSync('wrangler deploy --temporary', {
cwd: projectPath,
encoding: 'utf-8'
});
// Parse URLs from output
const deployUrl = output.match(/https:\/\/[^\s]+\.workers\.dev/)?.[0];
const claimUrl = output.match(/https:\/\/dash\.cloudflare\.com\/temporary-claim\/[^\s]+/)?.[0];
return { deployUrl, claimUrl };
}
The agent hands back both URLs. A human can visit the claim URL to take ownership, or let the deployment expire.
The 60-minute window is both the feature and the constraint. For deployments that need to survive without human intervention, temporary accounts are not the answer - you still need proper credentials.
Not every Cloudflare product is supported yet. R2 storage, Workers AI, and some enterprise features are absent from the launch list. Check the changelog for updates.
The claim flow requires a human with a browser. There is no API to convert a temporary account to permanent programmatically - someone needs to click through.
This is part of a pattern where infrastructure providers are adapting to agentic workloads. Stripe and WorkOS have similar agent-friendly provisioning flows. The thesis: if agents are going to deploy production infrastructure, the infrastructure needs to meet them halfway.
For developers building agents that touch deployment, this removes one of the harder integration problems. You no longer need to build OAuth handling, token management, or credential storage into your agent just to deploy a Worker.
Cloudflare has not published pricing specific to temporary accounts. Claimed accounts follow normal Workers pricing. Unclaimed deployments that expire do not appear to incur charges.
Each redeploy resets the timer. An agent actively iterating keeps the deployment alive. There is no explicit "extend" command.
It gets deleted with the rest of the temporary account when the 60 minutes expire.
No. Each wrangler deploy --temporary creates a new isolated account. Agents cannot collaborate on a single temporary deployment.
Yes. Your standard wrangler.toml configuration applies to temporary deployments. Bindings, routes, and other settings work as expected.
Yes. Temporary accounts do not require a paid Cloudflare plan.
The 60-minute expiration makes it unsuitable for persistent production workloads. Use it for prototyping, previews, and handoff flows - then claim the account for production.
Wrangler 4.x requires Node.js 18 or later.
Read next
Cloudflare's Agent Memory primitive. What it stores, latency profile, how it compares to mem0, and how to wire it into your stack.
9 min readAI agents use LLMs to complete multi-step tasks autonomously. Here is how they work and how to build them in TypeScript.
6 min readCoding changed more in the past two years than in the previous decade. We moved from manual typing to autocomplete, then to multi-file edits.
12 min readTechnical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.
Gives AI agents access to 250+ external tools (GitHub, Slack, Gmail, databases) with managed OAuth. Handles the auth and...
View ToolTypeScript-first AI agent framework. Agents, tools, memory, workflows, RAG, evals, tracing, MCP, and production deployme...
View ToolCDN, DNS, DDoS protection, and edge computing. Free tier handles most needs. This site uses Cloudflare for DNS and analy...
View ToolMulti-agent orchestration framework built on the OpenAI Agents SDK. Define agent roles, typed tools, and directional com...
View ToolPlan browser automation flows as inspectable product journeys before agents run them.
View AppDefine AI-assisted business automations without locking the workflow to one vendor.
View AppQueue and organize repeatable agent workflows before they become production automations.
View AppConfigure Claude Code for maximum productivity -- CLAUDE.md, sub-agents, MCP servers, and autonomous workflows.
AI AgentsStep-by-step guide to building an MCP server in TypeScript - from project setup to tool definitions, resource handling, testing, and deployment.
AI AgentsWhat MCP servers are, how they work, and how to build your own in 5 minutes.
AI Agents
Cloudflare's Agent Memory primitive. What it stores, latency profile, how it compares to mem0, and how to wire it into y...

AI agents use LLMs to complete multi-step tasks autonomously. Here is how they work and how to build them in TypeScript.

Coding changed more in the past two years than in the previous decade. We moved from manual typing to autocomplete, then...

Wire a Python LangGraph agent into a Next.js frontend using CopilotKit's co-agent architecture. Full walkthrough coverin...

The new wrangler deploy --temporary flag creates ephemeral Cloudflare accounts for AI agents. 60-minute deployments, no...

Hex's data-agent lab shows the practical eval pattern AI teams should copy: compare candidates against stable baselines,...

New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.