Printhouse

Your AI gets its own computer. You get a workspace to run everything with it.

[Printhouse @useprinthouse]

@blader We might just recognizing Claude's style because it's everywhere. You'd probably be able to recognize a different (human) writer's style too if it also was as ubiquitous

[Printhouse @useprinthouse]

Two ways primarily:
- The agent never sees real credentials, it sees fake keys which are swapped at the edge. Keys are safe even if the sandbox is compromised

- Users can set what network access their agent has, e.g. “all” / “connections only” / “none”. If a user selects “connections only” the blast radius is only what the user authorizes

More info on my approach here https://t.co/5lHCl3SiYI

[Printhouse @useprinthouse]

Hermes and OpenClaw are incredible tech, but they're still too unreliable and too operationally heavy for most non-technical users.

Printhouse is the same basic idea, an AI with a real computer, but packaged as a workspace people can actually use: no setup, better security, and a UI for seeing, steering, and working alongside the agent.

[Printhouse @useprinthouse]

@theo What are you doing for cloud agents?

[Printhouse @useprinthouse]

I built Printhouse to solve these pain points.

- no setup. just sign up and your bot is running
- credential proxy built in
- dedicated ui to show skills, connected apps, scheduled tasks, conversation threads, etc
- deeply integrated file explorer / viewer so it’s not a black box

[Printhouse @useprinthouse]

@mrhighfoster Amazing! Please let me know if you any issues or feedback that would make the product better for you

[Printhouse @useprinthouse]

I needed an AI that can run every part of my life, with a workspace where I can actually work alongside it, and security I can reason about.

So I'm building Printhouse.

https://t.co/b0w5m9UZat

[Printhouse @useprinthouse]

@honchodotdev Thanks. Currently my agent has no native memory, so I'm deciding whether to build that in addition to using honcho. Will experiment

[Printhouse @useprinthouse]

@mrhighfoster What are you hoping to use it for - or still figuring that out?

[Printhouse @useprinthouse]

@shafu0x Printhouse - effectively hermes or openclaw with lower technical bar, setup friction, and good security

[Printhouse @useprinthouse]

@ian_dot_so Feels like it happened already to me

[Printhouse @useprinthouse]

This works because extensions run on the same computer as your agent. Same filesystem, same connections, same credentials. Connected GitHub? Your extension can already use it. Connected Gmail? Same.

No deploy. No API keys to wire up. Just a panel that appears when you need it.

They're called Printhouse Extensions because they extend both your agent and the workspace UI https://t.co/9fLk0cvyE0

[Printhouse @useprinthouse]

People keep saying UIs will be ephemeral, generated on demand by AI. Nobody's actually shipped this.

I built something in Printhouse that I think cracks it.

[Printhouse @useprinthouse]

I told my agent I wanted a flashcard app for studying. It built one with a backend, database, and spaced repetition, and it showed up as a panel in my workspace.

Now I learn topics in chat, and when I want to solidify a concept, I tell the agent to add it as a flashcard. I quiz myself in the app later.

[Printhouse @useprinthouse]

I do not want to rely on Composio. Are there any alternatives that
- Give you access to tons of integrations
- Exposes a way to access with the underlying APIs themselves (not MCP tools)
- Don't force you to register the OAuth applications?

[Composio @composio]

We've identified a security incident that involved unauthorized access to certain internal Composio systems, impacting a limited number of customers.
We will share more as we learn more.

Please see our security bulletin:

https://t.co/y4WhykXc4B

[Printhouse @useprinthouse]

@PossibltyResult Check out printhouse, might fit your needs

[Printhouse @useprinthouse]

@nbaschez Actual coding in software engineering interviews feels totally performative these days

[Printhouse @useprinthouse]

I wrote more about the security model here - credential proxy, network access controls, and why user authority lives outside the agent runtime:

https://t.co/nk495fw1nA

[Printhouse @useprinthouse]

AI agents should not enforce their own permissions.

In Printhouse, the agent gets its own computer, but credentials and network access settings live outside that computer, where the agent can’t edit them.

It request changes, but it can’t rewrite the rules.