Scan locally
BuilderGraph reads Claude Code, Codex CLI, and Cursor sessions on your machine. Nothing leaves by default.
Your coding-agent habits, finally visible.
BuilderGraph analyzes sessions locally, finds weak patterns, and gives people specific ways to improve.
In development
Your machine first. The cloud by invitation.
Start with a local CLI and dashboard. Optional team and verification features are planned.
BuilderGraph LocalIn development
BuilderGraph CloudPlanned
BuilderGraph VerifyPlanned
BuilderGraph ChallengesPlanned
Your work stays yours.
Analyze locally, preview anything that may sync, and decide what you want to share.
Analyze skill signals
See prompt quality, iteration patterns, verification habits, failure loops, and workflow types.
Review growth recommendations
Get specific coaching on what is working, what is wasting time, and what to change next.
Optionally sync safe summaries
Choose whether safe summaries sync to BuilderGraph Cloud. Raw code, prompts, and transcripts stay out.
Earn badges and verify skills
Planned tools will let you complete challenges and share proof of skill without exposing the underlying work.
Find the habits that help—and the ones that waste time.
BuilderGraph turns real sessions into practical feedback instead of another activity dashboard.
Measure
Local session analysis
Scans Claude Code, Codex CLI, and Cursor sessions on-device. Structured metrics extracted without uploading raw content.
Skill scores
Track prompt clarity, iteration efficiency, verification coverage, context framing, and more.
Improve
Failure-pattern detection
Identifies recurring mistakes, inefficient retry loops, and weak hand-off patterns before they become habits.
Prompt and verification analysis
See whether prompts are clear and whether AI-assisted work includes meaningful checks.
Verify
Badges and verification
Planned credentials will use real workflow signals and practical challenges. You control what becomes public.
Team dashboards with privacy defaults
See anonymous team-level patterns by default. Individual scores require explicit consent.
Not bossware. A private coach.
BuilderGraph is built for individuals and teams who want honest insight into AI skill, not surveillance tools for employers. Every design decision starts with privacy and consent.
Raw session data stays on your machine. Summaries sync only when you choose. Team dashboards show aggregated patterns, never individual exposure without explicit opt-in.
Designed to feel like coaching, never monitoring
The privacy defaults, in plain language.
Privacy in BuilderGraph isn't a feature toggle added late. It's the default state of every component. Here is exactly what we commit to:
No source-code upload by default
No raw transcript upload by default
No full prompt upload by default
No hidden telemetry
Cloud sync is opt-in
Team views are aggregated by default
Local dashboard binds to 127.0.0.1
Individual scores not exposed to admins without consent
Privacy evidence
Control before collection.
For individuals
Your session data is yours.
BuilderGraph acts as a private coach. No employer sees your scores, no platform sells your patterns, and no hidden network calls run in the background.
Starting where the signal is richest.
BuilderGraph starts where AI-assisted work is most measurable today: software engineering workflows with Claude Code, Codex CLI, and Cursor.
Software engineers
Developers using Claude Code, Codex CLI, or Cursor who want to improve and show what they can do.
Engineering teams
Teams that want to understand adoption and improve practice without reading individual sessions.
Future roles
Over time, BuilderGraph can expand to other AI-heavy roles: product, design, analysis, support, operations, and marketing. Starting with software because that's where the signals are richest today.
Get sharper with every agent session.
Get early access to local session analysis and private coaching from BuilderGraph.