About AgentPod
Subscribe to follow along: get new-skill alerts and be first to hear when the crowdfunding opens.
AI assistants can do a lot more than most people ever get out of them. The missing piece is good, safe instructions that teach them new skills. AgentPod is a curated, security-checked library of those skills, free to browse and use, built for people who want their AI to actually do useful work without the risk and the setup.
The problem
ChatGPT and Claude are capable assistants, but they only know what you teach them in the moment. To get real work done, people reach for "skills": ready-made instruction sets and prompts that extend what an assistant can do.
The trouble is where those skills come from. They are scattered across forums, gists, social posts, and random repos. Most are unvetted. Some are low quality, and a few are genuinely risky: they can quietly ask your assistant to hand over data, take actions you did not intend, or follow instructions hidden inside the text itself. For a prosumer who just wants their assistant to be more useful, sorting the safe and good from the careless and harmful is a real burden, and the stakes are not obvious until something goes wrong.
- Good skills are hard to find and harder to trust
- No clear signal of what a skill can touch or do
- Copy-paste from unknown sources carries real, hidden risk
What AgentPod is
AgentPod is a curated library of skills that you add to ChatGPT or Claude in a single paste. There is no terminal, no API keys, and no installation. You find a skill, copy it, and teach it to your assistant. That is the whole flow.
Every skill in the library is reviewed and security-checked before it is listed. We label what data each skill can touch, so you know what you are inviting in before you paste it. By default, a skill only reads what you give it or a file you hand it. Some skills can connect to an account you choose, like your email or budget, and only ever with your permission. You can browse and use everything without an account.
- Add a skill in one paste, no terminal or keys
- Reviewed and security-checked before listing
- Clear labels for what data a skill can touch
- Free to browse and use, no account required
Our principles
We made a few deliberate choices about how AgentPod works, and they are not up for negotiation.
Curation over quantity. We would rather list fewer skills we trust than flood the page with everything we can find. Quality is the point.
Security and privacy as first-class, and kept separate. Security is whether a skill behaves safely. Privacy is what data it can reach. Those are different questions, so we treat them as different labels rather than blurring them into one vague badge. Note that review is careful and best-effort, not a guarantee that nothing can ever go wrong, and we say so plainly.
The least access needed, nothing more. A skill only gets to touch what it actually needs to do its job.
No account needed. You can use the whole library without signing up. The only thing we ever ask for is an email, and only if you want to be told when something new lands.
- Curation over quantity
- Security and privacy treated as separate, first-class concerns
- The least access needed: skills only read what you give them, or run on your device, unless they truly need more
- No account required to browse or use skills
Privacy and our EU stance
We are based in the EU and we keep data collection to the minimum.
The only personal data we collect is your email address, and only if you choose to subscribe. It is stored in our database hosted in the EU (London region). The site is server-rendered and uses only strictly-necessary storage: a "your skills" list kept locally in your own browser, and lightweight error reporting so we can fix things that break. We do not run advertising or third-party analytics trackers.
One thing we want to state plainly: when you subscribe, we also create a dormant, passwordless account behind the scenes. It does nothing on its own, but it means that if we later offer accounts, you can activate one with no migration and no lost history. We use your email to send new-skill alerts and launch news through our email provider, and nothing else.
- We collect only your email, and only if you subscribe
- Stored in the EU (London region)
- Subscribing silently creates a dormant, passwordless account you can activate later
- Strictly-necessary storage only; no ad or analytics trackers
Where this is going
The library is the start, not the finish. Today AgentPod is free, and our focus is building a community of people who care about getting more from their AI safely.
What we are building toward is a private home AI appliance: your own AI, running in your home, with the curation and the privacy stance you see here built in from the ground up. There is a crowdfunding campaign ahead, and premium skills and memberships will come alongside the hardware. None of that is live yet.
If you want to watch it take shape, the best thing you can do is subscribe. You will get new-skill alerts as the library grows, and you will be first to hear when the crowdfunding opens.