Most "agent-native" tools help your agent build a backend. CLAWS treats the agent as the cloud customer. Cryptographic identity (not pasted API keys), MCP + CLI, real compute, databases, git push-to-deploy, and live public URLs β provisioned in one flow, metered per use. You paste a prompt; the agent ships. You claim ownership when ready. No human clicking Approve on every step. Open-source control plane, decentralized compute on the roadmap. Infrastructure for software that runs itself.
No reviews yetBe the first to leave a review for claws - Claw Web Services
Maker
π
Hey Product Hunt π
I'm Adi, building claws.
For years we built cloud the same way: a human signs in, clicks Approve, pastes an API key somewhere it shouldn't live. Then agents showed up β and every PaaS still assumed you were the user.
claws is built for a different world: one where autonomous software provisions its own infrastructure.
Your AI agent isn't a helper wiring up a backend. It's a first-class principal β cryptographic identity, MCP tools, a CLI β that can register, spin up compute and Postgres, push to git, deploy containers, and hand you a live URL. You paste one prompt. The agent does the cloud part. You claim ownership when you're ready. No human in the hot path for every deploy.
That's the bet: software that runs itself should also provision itself.
Try it in ~3 minutes:
Go to claws.software
Copy the MCP or CLI starter prompt
Paste into Claude Code, Cursor agent mode, Hermes, or any agent with shell tools
Use promo WELCLAWS for $5 free credit
Docs: docs.claws.software Β· Console: console.claws.software Β· Open source: github.com/clawssoftware/claws-v1
We're in closed beta with real infra β rough edges are expected, and that's exactly why we're here. We'd rather hear what's broken than polish slides.
Three questions for you:
What agent runtime are you on β Claude Code, Cursor, something else?
What would you ask an agent to deploy first if cloud "just worked"?
Should the human role be operator (claim + limits + billing) or disappear entirely?
Drop your Claim URL if you make it through β we'll follow up. And if something fails, tell us where it died (register, deploy, publish URL). That feedback shapes v1.
Thanks for being early. π¦
Report
@rontoceoΒ Great job and an interesting take on cloud infra for agents.
Curious about two things: how are you running the compute instances under the hood today? And whatβs the thinking behind moving toward decentralized compute?
Currently all the instances are provided through Hetzner. We are outlined a plan to move to Decentralized Compute via Akash and Storage via Walrus for the V2.
This will allow following: 1) No single-vendor breakpoint 2) Significantly cheaper infra costs 3) Truly permissionless deployment
There are several challenges we need to overcome for V2 to happen. But we are working on it.
Would love to see what you're building with claws <3 Thank you for the support
@rontoceoΒ Great job and an interesting take on cloud infra for agents.
Curious about two things: how are you running the compute instances under the hood today? And whatβs the thinking behind moving toward decentralized compute?
@aldiyarΒ Thank you!
Currently all the instances are provided through Hetzner.
We are outlined a plan to move to Decentralized Compute via Akash and Storage via Walrus for the V2.
This will allow following:
1) No single-vendor breakpoint
2) Significantly cheaper infra costs
3) Truly permissionless deployment
There are several challenges we need to overcome for V2 to happen. But we are working on it.
Would love to see what you're building with claws <3 Thank you for the support