SignalGen
GitHubOpen dashboard
Google Cloud Rapid Agent Hackathon

From customer
signal to
product PR.

SignalGen is an event-driven and periodic AI product-iteration agent that turns feedback from your social media / customer channels into safe, reviewable product changes — keeping founders in control of every step.

Open dashboardView workflow
Illustrative Signal Room

Users need clearer onboarding

Sample UI only: this card shows how a signal, evidence, and pipeline feel in the product.

74%example
Evidence format
“I got confused during setup.”
0 mentionssample severity74% conf
“Where do I even start?”
0 mentionssample severity74% conf
Recommended change format

Draft a reviewable improvement for your product's getting-started flow, then wait for founder approval before any code changes ship.

Upload
Extract
Signal
Plan
Approve
Branch
Checks
PR
Preview
Agent workflow

Human-approved automation,
with guardrails.

The signal travels one visible loop — and never crosses the approval gate without you. Each run is stored, closing back into the next product iteration.

Illustrative product iteration loop
sample flow
intake
Feedback
screenshots → comments
74%
detect
Signal
confidence
draft
Plan
files · guardrails
gate
approval gate
Approve
human in control
ship
Safe PR
branch · checks · preview
stored
Memory
informs the next iteration
stored end-to-end ↺
feeds the next loop
active done approval gate · nothing ships without you
Step 1

Collect customer signals

Connect your social media / customer channels or upload feedback screenshots so SignalGen can watch for useful product evidence.

1
Step 2

Detect the product signal

Event-driven and periodic runs extract comments, cluster repeated pain points, and explain the strongest signal with evidence.

2
Step 3

Approve the implementation plan

Review the proposed product change, affected files, acceptance criteria, and guardrails before code changes happen.

3
Step 4

Create a safe PR

The agent edits your product repo on a branch, runs verification, opens a reviewable GitHub PR, and links real previews only when available.

4
MongoDB memory

The memory layer of the founder's product iteration loop.

SignalGen stores the path from customer evidence to founder approval to reviewable change, so future product decisions can learn from earlier runs.

Evidence
loop
Approval
gate
Trace
memory
01Original feedback source and extracted customer comments
02Top signal, confidence, evidence, and founder decision
03Generated plan, guardrails, and approval history
04Changed files, verification result, PR, and preview link when real