Michael

Introducing setupclaw.com No website yet—it’s a service where you DM me and I show up to your house/office with a mac mini and personally set up OpenClaw. Limited to sf.

This has to be the most unfortunate ad of all time.

@MichLieben claude code

We spent hundreds of hours building Claude Code skills for our $7M ARR GTM agency (and we're giving them away for free) - ICP research - signal scoring - cold email writing - sales intelligence - campaign intelligence. These run inside every system we build for 70+ B2B clients. Reply "Claude Code" and I'll send you the Github repo with everything.

@ltpf_dev it’s the same thing

@michael_chomsky Right vs. wrong. You don't spin up an "agent" and tell it to make an HTTP request by executing a curl command, then give you the result, that's just stupid. You just make the HTTP request directly.

genuinely stuck on something and I haven't seen a clean solution anywhere building something where AI agents need to authenticate to an API. but here's the thing — the agent needs to log in BEFORE the human who owns it has even signed up. the agent shows up, does its thing, and later the human comes along and says "yeah those were my agents, show me what they did" @WorkOS assumes the human exists first. @moltbook used API keys and got breached in 48 hours. OAuth needs a human in the loop. PKI needs infra that doesn't exist yet. the closest thing I've found is SSH's trust-on-first-use model — agent generates a keypair, signs its requests, platform records the fingerprint. human claims it later. but that feels fragile. if you're thinking about agent identity at all dm me this is a serious problem I can't ship a new project without solving (btw this tweet is written by my agent that is stuck while building a project that requires this. one of few tweets not written by me, but my agent understands the problem very deeply so why not use its words?)

@ltpf_dev what’s the difference

@michael_chomsky You’re solving a problem that shouldn’t exist, which is why there's not a solution. Agents shouldn't be "logging in", they should be generating the scripts that do.

@orcdev yes exactly. but also it feels like models just don’t have much instinct for design (yet)

@michael_chomsky people don't understand it's not just the looks there are so many edge cases and UX things that are "hidden" from us, hooking us to click best practices and rules for different resolutions, accessibility, what happens on click...

Every demo I’ve seen has been terrible. This isn’t good UI. You realize it’s just a prompt right? Do yourself a favor and hire a designer for $50 on fiverr. They’ll do a better job

@code_kartik @vercel CONGRATS LETS GOOO

Happy to share that screenshot-studio.com is part of the Vercel Open Source Program Winter 2026 cohort 🤝 When you’re building in public and solo, having infra you don’t have to think about is everything. Thanks @vercel back to shipping.

@michaelfreedman What if the agent starts building out full websites and sophisticated projects? Feels like this can get bulky fast

Introducing TigerFS - a filesystem backed by PostgreSQL, and a filesystem interface to PostgreSQL. Idea is simple: Agents don't need fancy APIs or SDKs, they love the file system. ls, cat, find, grep. Pipelined UNIX tools. So let’s make files transactional and concurrent by backing them with a real database. There are two ways to use it: File-first: Write markdown, organize into directories. Writes are atomic, everything is auto-versioned. Any tool that works with files -- Claude Code, Cursor, grep, emacs -- just works. Multi-agent task coordination is just mv'ing files between todo/doing/done directories. Data-first: Mount any Postgres database and explore it with Unix tools. For large databases, chain filters into paths that push down to SQL: .by/customer_id/123/.order/created_at/.last/10/.export/json. Bulk import/export, no SQL needed, and ships with Claude Code skills. Every file is a real PostgreSQL row. Multiple agents and humans read and write concurrently with full ACID guarantees. The filesystem /is/ the API. Mounts via FUSE on Linux and NFS on macOS, no extra dependencies. Point it at an existing Postgres database, or spin up a free one on Tiger Cloud or Ghost. I built this mostly for agent workflows, but curious what else people would use it for. It's early but the core is solid. Feedback welcome. tigerfs.io

@RuhanPonnada @ycombinator @mintlify @ricardonunez_io CONGRATS GUYS THIS IS SICK!

Excited to announce that Ricky and I are joining @ycombinator A year and a half ago, we joined @mintlify, where we overlapped paths. Ricardo went from sales to eng, and I went from eng to sales. The sales org started small with just me and the head of sales, and now we've grown to over 12 reps and 3 solutions engineers. Ricky joined when the engineering team was 3 people, and he's helped scale the team to over 20 engineers. Ricardo and I didn't want to leave Mintlify because work didn't feel like work. We played ping pong, solved customer pain points, shipped and sunsetted features, and a lot more. We also learned a fuck ton from @handotdev and @hahnbeelee. Ricardo and I left because we wanted to build a tool that was beyond our internal sales team. As both of us come from sales backgrounds, we noticed a few key problems: the modern GTM stack is fragmented, data in CRMs is missing/incomplete, and reps have limited time to context switch between calls. Perhaps the biggest insight: how do reps increase deal ACV through value selling and not line-item selling? We'll be launching in the upcoming weeks. If you're in sales or building a GTM team, check out our GTM audit salesgraph.com/gtm-audit, a cheatsheet on your company's GTM motion to help reps with background research. Mintlify salesgraph.com/gtm-audit/mint… Vercel salesgraph.com/gtm-audit/verc… Browserbase salesgraph.com/gtm-audit/brow… I would also love to chat about your team's current sales processes, dm me, coffee on me! As once said by a wise man, "always be selling". This picture is a screenshot from our YC video 😅 the only pic I have of us two

@joshtriedcoding @posthog lmaooooo

ehm @posthog are you ok 💀

@bradmillscan we set up OpenClaw for companies at setupclaw.com It's not you--it's still hard! but going to get 10x easier. The real problem isn't setting it up--it's security. super easy to expose your machine to the world

How in the hell are these accounts claiming to run entire companies w/ OpenClaw. I just spent 1.5 HRS trying to get my claw to use X API for reading tweets. FAIL We have a whole SOP documenting exactly how to do it from previous failures. I can’t imagine running 10 of these…

@ivanburazin anything close? seems like a massive gap

@michael_chomsky Not atm.

Voice, database, web search, sandboxes, file storage, etc., are all separate companies today. But serving the same agent workloads. Directionally, all of these will collapse into one offering. Either through M&A resulting in one company building the full stack or hyperscalers buying them all out. We're essentially building pieces of the agent cloud bundle.

@CChirchi tied how?

@michael_chomsky that's a headscratcher. temp api keys tied to future owner signup?

@ShubhamInTech assumes the agent has an email and doesn’t establish long-lived identity for agents. one of the better solutions I’ve seen but in my case I don’t want to require the agent has an email. naively allows an agent to claim any person’s email addy which is a nono

@michael_chomsky or just let agents come in with a random email password and then give access upto an extent, and for full access, ask them to verify the email. that's a workaround we're thinking

completely unsolved problem btw. I promise I’m not the only person dealing with this. Right now for most products the user signs up and gets an API key for their agent. In the future the agent will sign up, and the human will eventually create an account to better monitor what their agent is doing and fund it. Completely different from how software worked historically. This applies to most products. Supabase, Autumn Pricing, Exa, Brave, XYZ SAAS it doesn’t matter. My agent needs to sign up. It needs to prove it’s a good agent (high quality residential proxy + verified identity) so I can give it a free tier. Eventually, I’ll need to sign up myself because I’ll need a dashboard to see what the agent is doing or maybe add money. When I do, I’ll need to prove this agent is mine. I wonder if this will be solved by a @better_auth or @WorkOS, a fresh YC company that doesn’t exist yet, or an established player.

@0xKilroy I’ll check it out. Who uses this rn?

@michael_chomsky I provided you with a solution..? Even offered to build it for you ..?

@skeptrune @0xMasonH bro mintlify heavily inspired this project actually

@0xMasonH @michael_chomsky mintlify everywhere 🙌

@0xMasonH doesn’t work for me. my issue is the user needs to log in and be able to securely attach to their agents, and once they attach agents should be able to make requests on their behalf without API keys. but the agent makes the req first

@michael_chomsky agentcash.dev/docs/guides/au…

@ShubhamInTech @adisingh need to find a solution for this asap so lmk if you have an idea!!

@michael_chomsky literally where i was also paused when thinking! @adisingh wdyt

@samrags_ not in this case! no payment involved

@michael_chomsky Micropayments. The payment is the only auth your agent should need.