Decawork

This is a real problem I’ve been seeing with startups - someone on the team builds agents that run on personal keys, with no audit trail and everything is all over the place with no central identity or control.

Super interesting! Philosophically in alignment with you about having an AI stack that isn't locked into one provider.One question, since often security is part technical and part behavioral: what steps of behavioral change do employees need to take for this to work?

Congrats on YC Sarthak.. great launch.. good wishes for its success.. I remember working on AIM system which has similar need during RPA deployments.. great to learn about all your other startups too.. good luck..

Centralizing control and access permissions for internal AI agents makes complete sense as teams scale. Congrats on the launch!

The interesting question foe me is what happens when an amployees leaves the company . If their agents can be transferred and managed like company assets, that could be a huge operational win.

Access control is probably going one of the biggest challenges of the agent era. Having everything managed from one place a lot of sense.

Connecting AI spend directly to roadmap items makes the numbers much easier to understand. I like that approach.

Security plus visibility is probably where internal agents will need the most help. Curious how you balance storng controls with developer flexibility.

I’d be interested in seeing audit logs for agent activity, especially for teams handling customer or financial data.

I like the employee-style lifecycle for agents. Having a clear way to retire old agents could prevent a lot of security headaches.

The idea of putting employee-built agents under IT control feels especially useful as more teams start making their own tools. 👀

Treating internal AI agents like company assets makes a lot of sense. I like the focus on access and retirement instead of just deployment.

Good luck on the launch day!

Hey Product Hunt, I'm Sarthak - co-founder of Decawork (YC S26).AI coding tools made it easy for anyone to build an internal agent. The hard part starts when a teammate wants to use it. Now the agent needs company credentials, access to real systems, an owner, approvals, logs, maintenance, and a way for IT to stop it.Decawork handles that handoff. Your team still owns the job. Bring us the repo from Claude Code, Codex, Cursor, or whatever your team used. Decawork operates the agent with a company identity. IT decides what it can access, reviews sensitive actions, sees what it did, and can pause or retire it from one place.We think internal AI is unlocked by IT, not by one more agent builder. IT should not have to choose between blocking every experiment and rebuilding each one by hand.We've spent years on both sides of this handoff. I built AI systems at NVIDIA that shipped to OpenAI and Meta, then enterprise agents for Microsoft and Hitachi. Aman (my co-founder) built Barclays' AI compliance platform, led a consumer AI product to 100K+ monthly users, and founded a fintech serving 52K families. We'd love feedback from people building internal agents, especially the IT and security teams asked to let them touch company systems. What was the first thing that broke when your agent moved beyond its original builder?Find a time at decawork.ai or reach out at sarthak@decawork.ai.