Skip to content

Forge

Forge is a product-first agent system that helps people discover, evaluate, and build product opportunities.

Code Demo Cerebral Valley

Core trick

Ranks product opportunities from repo signals and critique agents, then opens a PR from an approved build brief.

Tags

productivitydev toolsagents

Description

Forge is a product-first agent system that helps people discover, evaluate, and build product opportunities. It supports both connected products, like an existing GitHub repo, and new-product ideas. The system ingests project context, manual ideas, repo signals, public research, and user preferences, then ranks opportunities, runs critique and Bull/Bear review, recommends a next action, and lets the user approve a direction for prototype or MVP generation through the dashboard. It then autonomously builds it for them and opens a PR so the demo can actually be tested.
nadavital/Forge★ 00 forksTypeScriptCreated May 23, 2026 · last push May 26, 2026

README excerpt

# Forge

Forge is a product-first agent system for turning project context, source evidence, and user taste into reviewed product opportunities and approved prototype or MVP build work.

The current repository is an early vertical slice, not just a scaffold. It includes a Next.js dashboard, local/Supabase storage contracts, connected-repo discovery, semantic repo analysis through a Gemini/Antigravity-style agent path, opportunity review actions, build brief generation, simulated and managed builder adapters, and a simple reflection proposal loop.

Forge is still not a complete autonomous product-discovery platform. The strongest current path is connected-product review from a GitHub repo. New-product mode captures context and signals, but it does not yet create agent-backed recommendations without a real discovery/research step.

## Current Product Shape

Forge currently behaves like a product review and build-orchestration workbench:

1. Create a connected-product or new-product project.
2. Store explicit product context, source configs, triggers, and user preferences.
3. Run Dream for a project.
4. For connected repos, Forge fetches GitHub repo metadata, README, issues, and tree context.
5. A repo analysis agent inspects the connected repo and returns project knowledge plus evidence-backed opportunities.
6. Forge validates and stores signals, opportunities, evaluations, prototype records, and pipeline run metadata.
7. The dashboard renders a product review: recommendation c

Team