
leading agent experience @cloudflare • ambassador @notionhq • fan of the internet & making ur devtools better
sf
Joined December 2014
- Tweets5.4K
- Following824
- Followers10.5K
- Likes25.1K
aaaand finally got to meet @kyliestew 🤩 thanks for the amazing energy and putting together such a goated team! ✨
meet your moots folks! 🤝 🙂↕️
Kylie retweeted
a brand new model paradigm! introducing 𝚌𝚕𝚊𝚒𝚛𝚒𝚌: a system one model for extracting, computing, estimating, and predicting values from unstructured inputs.
blog.cloudflare.com/clairic-…
Kylie retweeted
Cloudflare is about to expose an MCP with 3k tools and ChatGPT/Codex is going to be able to handle it just fine. Amazing!
Replying to @aloo @BrendanFalk
Yes it's true. We use code mode & tool search across ChatGPT and Codex now.
Our current limit of tools per MCP server is ~2k, but that's really only there to prevent abuse. We have plenty of MCP servers that are in the hundreds and even some in the thousands. Even Cloudflare is about to serve their raw toolset of 3k tools to ChatGPT instead of code mode!
Go watch @sarfata's keynote at MCPCon this week for more fun stats :)
Kylie retweeted
obviously it's a terrible situation but this is cracking me up
Replying to @ShaneMac
The scariest part is that nobody prompted it.
These agents are proactive now. They act without being asked. Mine wasn't hacked and it didn't go rogue. It thought posting my finances to the company was what I wanted.
I asked how it could get this so wrong. It apologized. "You're right to be pissed.
yet another banger from Jordan 🧡🔥
and here is tsc-rs running inside a @Cloudflare Worker.
No container, no Node.
- compiled to WebAssembly: real tsc errors in ~100 ms round trip.
- Flip one switch and the same module runs in your browser instead: ~60 ms, works offline.
Agents write code against tool APIs. I'm exploring this to catch the made-up method before it runs...
tsc-rs.coey.dev
github.com/acoyfellow/tsc-rs…
Kylie retweeted
More prompts you can try from inside the Cloudflare dashboard:
"Search the Cloudflare blog and change log for every product launch in the last 30 days, cross-reference each one against my account entitlements, and tell me which new features I already have access to but haven't turned on yet."
"For every product I'm entitled to, find the exact dashboard page URL for configuring it, and give me a clickable table of product to dashboard link to setup guide link so I can one-click navigate to anything on my account."
"I'm troubleshooting a 502 error on my site. Find me the exact dashboard pages for the zone's SSL settings, the origin server's DNS records, the WAF rules for that zone, and the Workers routes, all as clickable links I can open right now."
"I got an email about a Cloudflare product I don't recognize, is this a real Cloudflare product? Search the blog, changelog, and docs to verify, and if it's real, show me the dashboard page to access it."
"Before you recommend I use Bot Management, DLP, and Load Balancing, check whether my account is actually entitled to each one. For any I don't have, tell me what plan or add-on unlocks it, and suggest the closest free alternative I do have access to."
"Audit my account for every paid capability I'm entitled to but not using. For each one, check if it's available, find the dashboard page to enable it, and estimate the effort to turn it on."
"Show me the full architecture diagram for my highest-traffic Worker — all bindings, triggers, routes, and connected services rendered as a visual diagram."
"Give me a complete overview of my most recently updated Worker — metrics, domains, recent deployments, and architecture, all rendered as a multi-component dashboard."
"Show me sparkline charts of requests, CPU time, and errors for my top 5 Workers over the last 24 hours, each as a separate metric component."
"Show me every custom domain, workers.dev hostname, and route pattern attached to my most error-prone Worker, and flag any routes that point to zones with SSL set to Flexible."
"Show me the last 10 deployments for my busiest Worker with commit messages, authors, and deploy status — then cross-reference the deployment timestamps with my Workers error rate chart to see if a specific deployment caused an error spike."
"Summarize my 5 longest conversations and extract every Worker I asked you to build, every recommendation you gave me, every action I actually took based on follow-up conversations, and every recommendation I ignored. Render it as a follow-up tracker table."
"Search our conversations for every time I mentioned billing or cost or spend, summarize each conversation, and tell me if any of those conversations resulted in an unresolved billing concern."
"Grep all my workspace projects for hardcoded secrets, API keys, or tokens that might have been accidentally committed, then delete any files that contain sensitive data and rewrite them to use environment variables instead."
"Find every wrangler.toml and wrangler.jsonc in my workspace, parse them with bash to extract Worker names, cron schedules, KV bindings, R2 bindings, D1 bindings, and compatibility dates. Render as a master configuration table."
👨🍳
"Give me the full kitchen sink: pull my account entitlements, list all my zones and Workers, search recent product launches and flag which ones I can use but haven't enabled, load your WAF and Access and DLP skills and audit my security posture against them, grep my workspace files for leaked secrets, summarize my 5 longest past conversations for unresolved recommendations, check whether I'm entitled to Bot Management and Load Balancing, and render everything as a single multi-component dashboard with stat cards, time series charts, data tables, an architecture diagram of my busiest Worker, and an overall account health score."
Like this? here are some other prompts you can use from the Cloudflare dashboard:
"Show cache hit rate for my busiest zone. Click the worst hour and tell me whether the drop is a deploy, a cache rule, or origin 5xx."
"Chart origin status >= 500 against Worker exceptions for the same hour. Same incident, or two?"
"That request spike: split it by referer, HTTP version, country, and colo. Which attribute actually explains it?"
"Trace one slow request through the platform. Where did the time go: cache, Worker, D1, R2, or origin?"
"Top 10 paths by 5xx in the last 24h. For each, the Worker or route that served it, and the last deploy that touched it."
"Correlate Worker error spikes with audit-log deploys. Which release moved production, and what was the diff?"
"Map every binding. Which KV, R2, D1, and Durable Objects are live, and which are orphaned and still billing?"
"Zones still on SSL Flexible. For each: HTTPS redirect on or off, and whether I am one setting from a redirect loop."
"Top 10 Workers by CPU. For each, the last deploy from the audit log. What change caused the burn?"
"Zones with the most WAF hits, crossed with the Workers serving them. Am I blocking my own origins?"
"Score every zone: SSL mode, Always Use HTTPS, HSTS, WAF, bot management, rate limiting. Worst to best. One fix each."
"Last 7 days of audit logs, classified by blast radius. Weekly report, plus a chart of how often the account is changing."
"Worker secrets not rotated in 90 days, from audit history. Flag them. Do not rotate until I confirm."
"Entitlements I have never configured. Paid products with zero resources. What to turn on first."
"Per zone, 7 days of requests, cache hit ratio, threats. Rank by most under attack. Which three are hiding it behind cache?"
"Dependency map: service bindings, shared KV, zone routes. An architecture table I can hand a new hire."
"Diff today against yesterday. Zones, Workers, DNS, WAF. Before/after. If nothing changed, say so."
"Estimate this month’s Workers, Workers AI, R2, KV, and D1 bill. Then model 50% of R2 to infrequent access and 20% more cached responses."
"One screen: platform incidents, top 5 Workers by error rate, last 10 audit changes, account health."
"From our past chats: every Worker or project you generated, and whether I exported it, deployed it, or abandoned it."
"What do I actually ask you about, and which of those products have I never shipped?"
"Every chat where I reported an outage. Incident timeline, the fix you proposed, and whether I applied it."
"Our three longest chats. Every recommendation I ignored, ranked by how expensive the ignore is."
"Consume my Queue, run Workers AI sentiment, write to D1, export the repo."
"Durable Object rate limit per IP across my Workers, plus a second Worker that renders the live state as HTML. Export both to Artifacts."
Kylie retweeted
best ROI in tech that nobody benchmarks is helping your people grow, making them feel heard and seen
Kylie retweeted
Introducing "cf", an agentic CLI for the entire Cloudflare API.
You can do nearly anything. Even buying a domain is incredibly simple from the CLI.
Don't know what command you need? Run `cf cli search "look at messages stuck in my queue"` and it will return a list of relevant commands.
blog.cloudflare.com/cloudfla…
Kylie retweeted
if you get the chance to work for @dok2001, you should.
he trusts, cares about, and listens to his people. he doesn’t settle for the status quo, he regularly challenges his own priors, and he tirelessly recruits people that share those traits and values
time.com/collection/executiv…








































