Deploy

tin deploy app.tin uploads the compiled program and answers with a live address. No account, no config file, no container.

$ tin deploy hello.tin
uploading hello.tin (3 KB) to https://api.tin.ad

live: https://eager-wombat-a7tp.tin.ad   anonymous, expires in 24 h (tin login keeps it)
  agent Hello: curl -N https://eager-wombat-a7tp.tin.ad/agents/Hello -H 'accept: text/event-stream' -d '{"prompt":"hello"}'
    EIP-8004: https://eager-wombat-a7tp.tin.ad/agents/Hello/registration.json

The output format of tin deploy, with the name from the recorded run on 16 November; every deploy gets its own name.

What each agent gets

EndpointWhat it is
POST /agents/<Name>JSON answer, or a token stream with accept: text/event-stream
/agents/<Name>/wsWebSocket: one conversation per socket
POST /mcpthe program's tools for any MCP client
/agents/<Name>/registration.jsonan EIP-8004 registration file; on-chain registration is opt-in, test networks only

Routes you write in a server block are served the same way, with the request's input tainted until the program parses or approves it.

How deployments are kept apart

Price and limits

Free during the preview. There are no paid plans yet.

Anonymouslives 24 hours · 10 deploys per address per day · 1,000 requests per deployment per day · 60 per minute per address · 6 MB per upload
Signed in (tin login, GitHub)kept until you remove it · up to 20 per account · your own name with --name · tin ls, logs, rm, secrets from any machine

In a browser: sign in with GitHub · your deployments.

Other clouds

tin build --target cloudflare | vercel | cloudrun | azure | lambda | pages writes a deployable folder for your own account, and the build refuses what a target cannot run. Targets and what each supports.

Report abuse

$ curl -X POST https://api.tin.ad/v1/abuse -d '{"url":"https://name.tin.ad/","reason":"phishing"}'

A person reads every report.