OVERSEER
Initialising guardian layer...
System Status
Guardian Layer — Active
Build — 0.1.0-alpha
OVERSEER WATCH
Analyses today0
Issues caught0
Files watched0
LIVE
The Guardian Layer
OVERSEER
Watching what your AI builds — in real time
Install in 30 seconds
$ npx overseer watch
Free tier · No credit card required
Research
What's your biggest frustration with AI agents?
Your answer shapes what Overseer solves first.
Please write something first.
Thank you.
This shapes what we build.
SCROLL TO WITNESS
overseer — guardian layer
$ [system] initializing guardian layer...
$ [daemon] attaching to filesystem events on /src
$ [active] intercepting agent modifications in real-time
$ [audit] detected 400 lines generated by AI agent in 0.2 seconds
$ [alert] CRITICAL VULNERABILITY DETECTED in auth.js:5
$ [reason] hardcoded JWT secret — token can be forged bypass
$ [warn] missing rate limits on /api/login endpoint
$ [resolve] translating changes to plain English narrations...
$ [success] comprehension gap closed. you are fully aware of what was built.
$ [sync] committing context to permanent knowledge base
$ [ready] awaiting next human command or agent modification.
847 developers on the waitlist
65% of devs use AI tools weekly
41% of all code is AI-generated
#buildinpublic · indie hackers · product hunt
The Problem

AI codes faster than you can understand

01
45%
of developers say debugging AI-generated code takes longer than writing it themselves — because they never understood it in the first place.
02
45%
of the time, AI coding tools choose the insecure option. Hardcoded secrets. Open databases. Missing rate limits on auth endpoints.
03
2×
Code churn has doubled since AI agents became mainstream. Code rewritten within 2 weeks — GitClear, 153M lines analysed.
Competitive Landscape

Not a code reviewer.
Something entirely new.

Tool Type Timing Strength Weakness Why Overseer wins
Cursor
AI IDE
 AI-first IDE + agent During + post Fast multi-file edits, great UI You still manually review everything Adds live supervision on top of Cursor — catches what Cursor misses before you hit Accept
Claude Code
Agent
 Terminal autonomous agent After proposal Deep reasoning, runs tests Heavy copy-paste review loop Removes the review step entirely — watches Claude Code live instead of you reading diffs manually
GitHub Copilot
Autocomplete
 Inline autocomplete + Workspace Inline + PR only Seamless in VS Code, cheap Post-write or PR-only reviews Pre-accept and session-wide — Copilot doesn't give live analysis cards across all files
CodeRabbit
PR Review
 AI code review for PRs After PR Multi-platform PR reviews Only works after code is written and pushed Real-time during the coding session — not after you've already committed the problem
Cursor BugBot
Bug Hunter
 Bug hunter inside Cursor Post-edit Finds complex bugs Reactive, not persistent or always-on Always-on from line one — persistent cards, full session context, never needs to be triggered
Windsurf
Agent
 AI coding agent Generate → review Fast code generation Same manual review pain as everything else Turns the generate-review loop into hands-free live supervision with no manual step
Real-time / during session
After the fact
The Solution

Three components. One guardian layer.

01
The Daemon
Runs silently on your machine. Watches the file system directly. Detects every change your AI agent makes within milliseconds. Works with every IDE — no plugin required.
02
The Engine
Each change is sent to our analysis engine. Returns severity level, a plain-English title, and a 1–3 sentence explanation any developer can understand — instantly.
03
The Dashboard
A live feed beside your IDE. Like a sports commentary feed for your codebase. Every card shows what was built, whether it's safe, and what you need to know right now.
Pricing

Start free. Scale when you're ready.

Free
$0
forever
  • Unlimited analyses
  • 1 project
  • 7-day history
  • All AI agents supported
Starter
$7/mo
billed monthly
  • Unlimited analyses
  • 3 projects
  • 30-day history
  • All AI agents supported
Pro
$19/mo
billed monthly
  • Unlimited analyses
  • Unlimited projects
  • 90-day history
  • Full security reports
  • Priority processing
Team
$49/mo
up to 5 developers
  • Unlimited analyses
  • 5 developers
  • Unlimited history
  • Shared dashboard
  • Team security overview
The Founder
A
A note from the builder

I'm 17, self-taught across full stack, mobile, backend, and AI tooling — building Overseer solo from India with no funding.

I know enough to read the code my agents produce. That's actually what made the problem worse. I'd watch an agent tear through 400 lines, understand maybe 60% of the decisions it made, and ship it anyway because the deadline was real. The comprehension gap isn't about not knowing how to code — it's about not being able to keep up with a machine that never sleeps.

Overseer exists because even experienced developers can't audit everything an AI agent produces in real time. I built the tool I kept wishing existed.

Ashish Goswami · Founder · Follow the build →
Questions

Everything you want to know

Does it work with Cursor, Claude Code, and Copilot?
Yes — all of them. Overseer watches the file system directly, not the IDE. It works with every AI coding agent and every editor including VS Code, JetBrains, Cursor, and Windsurf. No plugin needed.
Is my code sent to your servers?
Only the diff — the lines that changed — are sent for analysis, never your full codebase. Diffs are processed and discarded. We don't store your code.
How do I install it?
One command: npx overseer watch. A browser tab opens automatically. Start your AI coding session and Overseer runs silently in the background.
When does it launch?
The MVP is 3–4 weeks away. Waitlist members get early access first, along with updates on the build as it happens.
Do I need to be technical to use it?
No. Overseer is designed specifically for non-technical founders and vibe coders. Every explanation is in plain English — no CS degree required to understand what your AI just built.
Is the free tier actually free?
Yes, genuinely free. Unlimited analyses, 1 project, 7-day history. No credit card required. Upgrade to Starter ($7) for more usage, Pro ($19) for the full feature set, or Team ($49) for shared dashboards across your whole team.
Early Access

Built by a vibe coder,
for vibe coders

17 years old. Solo founder. Building Overseer using Overseer.
The tool watches itself being built.