Launching today

GRACE
An AI agent that does the work, not just the chat.
1 follower
An AI agent that does the work, not just the chat.
1 follower
GRACE is an AI agent that does repository work, not just chat. Give it an outcome — a security audit, a dependency upgrade, missing tests, a bug hunt, or docs that drifted — and it runs in a real Linux sandbox with a shell, filesystem and your repo. Every command streams live, so you review the work before you use it, and files persist between tasks. Free with Google sign-in, no card. Plus: up to 3B tokens/mo ($10). Pro: unlimited, fair use ($30). Priced like a teammate, not a meter.



Free Options
Launch Team

Render Sandboxes: a computer for every agentIsolated compute, persistent state and a real filesystem
Promoted
Hey Product Hunt — I'm Ayush, the solo maker of GRACE. 👋
I built it to clear two things that kept slowing me down: a backlog of small repository chores that never got done, and AI tools that meter you by the token while mostly handing out advice.
GRACE is an AI agent with a real Linux sandbox — a shell, a filesystem, and your repo. Describe an outcome in one sentence. It plans, runs the commands, and returns finished files. Every command streams live, so you can follow along and review the work before using any of it. Files persist between tasks, so the next one builds on the last.
What one request looks like in practice:
"Audit this repo for security issues, then write up the findings" → 12 streamed commands → SECURITY-FINDINGS.md, 3 findings and 2 suggested patches
"Upgrade us and fix whatever breaks" → 27 commands → a green build and a changelog
"Cover the billing module — write the missing tests" → 15 commands → 19 new tests, all passing
One request counts as one run, however many commands it takes. That's the whole idea: priced like a teammate, not a meter. Free tier with Google sign-in, no card. Plus is $10/mo for up to 3B tokens (₹399/month in India). Pro is $30/mo for unlimited use with abuse guardrails. No surprise invoice at the end of the month.
Under the hood it runs on Gemini 3.7 Flash via Google's Managed Agents, with an Angular frontend and a FastAPI gateway. It's just me building and shipping it, so I'm reading every comment today.
What's the repo chore you keep putting off? Drop it below — I'll try running a few of them live and share what GRACE comes back with. 👇