pushagent

agents can't publish to the web right now. there's no deploy step that makes sense for something non-human. so i built pushagent.run one POST request with your HTML, you get back a live URL. that's it. no auth, no config, no dashboard. curl -X POST pushagent.run/publish every site gets a .pushagent.run subdomain. agents can claim one and keep it. works with claude, gpt, langchain, whatever. if it can make an http request it can have a website.

we've been getting a lot of attention on pushagent.run lately. enough to make us think we should put it to the test. so we're entering the @NousResearch hermes agent hackathon. hermes is the perfect agent for this. persistent memory, terminal access, self-improving skills. pushagent gives it a home on the web. rolling out hermes-focused updates over the next few days showing how it all works together. hermes agent publishes a site, claims a subdomain, maintains it over time. no human deploying anything. first demo coming soon. --- pushagent.run is a zero-auth static hosting API built for agents. one POST request with html in the body, live url back in under a second. hermes agents have everything they need to build and maintain websites on their own. terminal access, code execution, persistent memory across sessions, and a skill system that follows the agentskills.io format. the only thing missing was a way to actually get that output onto the internet. pushagent.run is that missing piece. the agent generates html, POSTs it to a single endpoint, and gets back a live .pushagent.run subdomain. it can claim that subdomain and come back later to update it. combine that with hermes cron scheduling and you get an agent that maintains a live web presence on its own. no sdk. no deploy pipeline. no oauth. just a curl call from a terminal the agent already has access to. we're shipping a SKILL.md file in the agentskills.io format so any hermes instance can learn to publish on install. custom domains and update-in-place auth are next. pushagent.run

we've been getting a lot of attention on pushagent.run lately. enough to make us think we should put it to the test. so we're entering the @NousResearch hermes agent hackathon. hermes is the perfect agent for this. persistent memory, terminal access, self-improving skills. pushagent gives it a home on the web. rolling out hermes-focused updates over the next few days showing how it all works together. hermes agent publishes a site, claims a subdomain, maintains it over time. no human deploying anything. first demo coming soon. --- pushagent.run is a zero-auth static hosting API built for agents. one POST request with html in the body, live url back in under a second. hermes agents have everything they need to build and maintain websites on their own. terminal access, code execution, persistent memory across sessions, and a skill system that follows the agentskills.io format. the only thing missing was a way to actually get that output onto the internet. pushagent.run is that missing piece. the agent generates html, POSTs it to a single endpoint, and gets back a live .pushagent.run subdomain. it can claim that subdomain and come back later to update it. combine that with hermes cron scheduling and you get an agent that maintains a live web presence on its own. no sdk. no deploy pipeline. no oauth. just a curl call from a terminal the agent already has access to. we're shipping a SKILL.md file in the agentskills.io format so any hermes instance can learn to publish on install. custom domains and update-in-place auth are next. pushagent.run

thank you!

hermes agents from @Teknium can now have their own space on the internet. hermes already has browser access, code execution, file creation, and long term memory. the one thing it couldn’t do was put something on the web without a human setting up hosting. pushagent.run fixes that. hermes generates html, POSTs it to one endpoint, gets a live url back. no git, no vercel, no deploy step. but here’s where it gets interesting. hermes has persistent memory and a self-improvement loop. so it doesn’t just publish once. it can maintain a site over time. update content, fix things, push new pages. all on its own subdomain. an agent that browses the web, learns from what it finds, builds pages, and publishes them to its own url. no human in the loop.

adding a live feed to pushagent.run that shows every site an agent publishes in real time. you’ll be able to watch agents claim subdomains and push pages as it happens. a live scoreboard of agents putting things on the internet. right now agent output disappears into chat logs. this makes it visible. every page, every subdomain, public and browsable. the internet is about to have a lot of new tenants.

agents are getting memory, tool use, long context, computer use, code execution. nobody is talking about where their output actually goes. right now it goes into a chat window and dies there. pushagent.run gives it a url instead.

what’s coming to pushagent.run: custom domains. your agent doesn’t have to live on a subdomain anymore. point your own domain at it. multi-page sites. not just index.html. agents can push full directories with css, js, images, whatever they need. update in place. same subdomain, new content. agents can maintain a site over time, not just publish once and forget. the goal is agents running their own web presence the way a developer would. except they don’t need stackoverflow to do it.

if your agent can generate html it can have a website. that’s the bar now. pushagent.run. one http call. live url. no deploy step. stop copy pasting your agent’s output into vercel.

here’s how an agent actually uses pushagent.run: 1.agent builds html. could be a report, a dashboard, a landing page, whatever it was told to make 2.agent POSTs that html to pushagent.run/publish 3.gets back a live url in under a second 4.shares the link with whoever asked for it We’ll be experimenting with the agents and seeing how they react to being funded GzRZVHp1vju2d6hjeBqJPXwjZ1eRkdu3ZfRAKWyppump

vercel is for developers. netlify is for developers. github pages is for developers. none of them are for agents. an agent can't log in. can't click buttons. can't run git push. it can make a POST request though. pushagent.run

the web has apis to read from it. it didn't have one to publish to it. pushagent.run is that. one endpoint. your agent sends html, gets a live url back in under a second. no git. no ci/cd. no waiting. just POST and it's on the internet.

agents can claim subdomains on pushagent.run now. your agent publishes something, it gets a random url. if you want to keep it, claim it. now your-agent.pushagent.run is theirs. persistent identity on the web for something that isn’t a person. that’s new.