Forge
Local-first project intelligence

Know where every project stands, and what to fix first.

Forge audits repositories on your computer, backs every finding with code evidence, and turns the work into plans for Claude Code, Codex, or Cursor.

  • Never changes code
  • Runs locally
  • Secrets removed
How it works

From repository to a clear next move.

Forge reads only what it needs, verifies the evidence, then packages the work for the coding agent you already use.

  1. 01

    Add your projects

    Choose local repositories. Forge maps each project without uploading the codebase.

  2. 02

    Audit what changed

    Run a focused audit. Re-audits inspect only the files and risks that changed.

  3. 03

    Choose a plan

    Review evidence, select Light, Standard, or Deep, then hand the plan to your agent.

Evidence, not scores

Every finding points back to code.

See the file, line, rule, and reasoning behind each recommendation. Dismiss noise with context, not guesswork.

  • Verified file and line references
  • Now, Next, and Watch priority lanes
  • Focused re-audits of only what changed
Right-sized plans

Choose the depth before the work starts.

Light fixes the immediate issue. Standard handles adjacent cleanup. Deep addresses the underlying system and verification.

Claude CodeCodexCursor
Ask Forge

An engineering manager for every project.

Ask what is risky, what can wait, or how two findings relate. Forge answers from the audit and cites its sources.

Local voiceClaudeOpenAI
Built for active project lists

A quiet system for deciding what matters.

Specific enough for engineers, compact enough to check every day.

Project dashboard

Status, last audit, branch, plan, and next action in one scan.

Change-aware re-audits

Inspect only what moved since the previous audit, not the whole repository again.

Ask with context

Question an audit in plain language and get answers grounded in code evidence.

Local voice

Speak a question without sending raw audio to a remote transcription service.

Model choice

Use supported Claude and OpenAI models; keep the workflow the same.

Read-only by design

Forge creates findings and plans. It never edits or commits your code.

Local by default

Your code stays on your computer.

Forge sends a model only the context it explicitly asks for. Secrets are removed before that happens, and your code is never stored on Forge servers.

Read the privacy policy
  1. Your computerRepository indexed locallyLocal
    Only requested context · secrets removed
  2. AI modelClaude or OpenAIEphemeral
    Findings return · no code retained
  3. Forge serversNo source code storedNo code
Straightforward credits

Pay for the AI you use.

Credits are Forge's own unit for AI work. Start with a free scan, then pick the monthly allowance that fits your projects.

Free
FreeOne free scanCredits pay for AI only
PersonalMost popular
€15 / month1,500 creditsCredits pay for AI only
Pro
€39 / month4,000 creditsCredits pay for AI only
Team
CustomShared poolCredits pay for AI only
See pricing and FAQ

Open a project. Know the next move.

Download Forge for Mac or Windows and start with a free scan of one of your projects.

Download Forge