TrustClaw by Composio - Self-hosted AI agent that connects 1000+ apps on Vercel

Hi there PH. I'm back with another banger launch and this time an open-sourcing project.

A bit of context: I came across TrustClaw on Twitter, before I joined Composio. Every now and then I'd run into a founder who tells me they were using it as their daily driver.

Today Sarah(MTS at Composio and creator of TrustClaw) made it open source on GitHub. It is MIT-licensed, with a CLI that gets you from zero to a working instance in two minutes. The inner founder in me has been twitching about this all week and there's a real category of people who want their own always-on agent, on their own infra, with their own data, and not a hosted SaaS sitting between them and their accounts. TrustClaw is the cleanest version of that I've seen.

Huge shoutout to Sarah who shipped in less than 4hrs and turned this from an idea to what it is today.

Deploy a 24/7 agent via Vercel that talks to you on Telegram, remembers things, and plugs into ~1000 apps over OAuth via Composio, but didn't want to ship your tokens to a third party. Clone the repo and build your agent in < 2 mins. Btw OpenClaw but more secure.

Sarah is here to answer any questions. Let us know what you think. All feedback welcome :)

Hey, creator of TrustClaw here!

3 months ago OpenClaw had taken the internet by storm and I was immediately inspired to build some way to make it easier for non-technical people (and lazy people like me) to spin up their own personal agent without having to deploy their own docker container or set up plugins, skills, MCPs, CLIs for all their apps. So I built my own version and launched it and it went viral on X. After months of convincing, I am so excitid to announce today that it is open-source!

The craziest part is you can deploy all this in a single command!

npx @ComposioHQ/trustclaw deploy

Or, without touching any code at all via our Vercel template.

When preparing this project for the open-source community, easy deployment was a MUST. Vercel's stores (database + redis), cron jobs, AI Gateway, AI SDK covered absolutely everything this project needed. Their ecosystem is incredibly powerful and made this project so easy to get up an running.

Is there a way I could run this on my laptop? Because I would try a safer version of OpenClaw. I haven’t even tried OpenClaw yet because I think it’s not secure

Being able to self host is a big deal for personal agents that have access to my Telegram or Gmail. Having clear audit logs to see exactly what actions were taken while I was “asleep” would really help build trust in the system.

The 'doing things while you sleep' promise is bold! What kind of security guardrails are in place to ensure the AI doesn't perform unauthorized actions in a production environment?

vercel hobby caps functions at 5 minutes and cron jobs only run once a day. for anything beyond basic scheduled tasks that seems like it could hit a wall pretty fast. are most users running this on hobby or does it realistically need vercel pro to be useful day to day

The combination of scheduled actions, self hosting, and access to real apps through OAuth makes it easier for people actually building workflows they rely on daily, especially with the added control and privacy that comes from being able to run it yourself. Good job guys.

One click to deploy is insane, no more excuses to not ship. Can't wait to see what people build on top of this 👀

When the agent performs scheduled actions while you're asleep, what's the rollback mechanism if it makes a mistake? Can you audit what it did and undo bulk operations, or is it fire-and-forget once the cron triggers?