Auto-generate beautiful, agent-readable developer identity cards from GitHub.
No sign-up. No scraping. Paste a GitHub username -- get a visual card, structured JSON, AI-powered insights, and terminal output in seconds.
We optimize GitHub profiles for humans -- clean READMEs, pinned repos, contribution graphs. But AI agents are already evaluating developer capabilities, and they're scraping GitHub the same ad-hoc way we scraped resumes 10 years ago.
DevCard creates a structured protocol for developer identity. A devcard.json that says "I'm an ML engineer who ships production PyTorch, with 95% test coverage across my repos" -- readable by both humans and machines.
- 14+ signals extracted from public GitHub data (languages, stack, activity, code quality, collaboration style)
- Dual scoring -- Human Visibility (recruiters) + Agent Readiness (AI tools)
- MCP server -- AI agents query developer profiles as structured data, not scraped HTML
pip install devcard
# (Recommended) Set your GitHub token for 5000 req/hr instead of 60
export GITHUB_TOKEN=$(gh auth token)
# Generate your card
devcard generate YOUR_USERNAME
# Get an SVG for your GitHub README
devcard generate YOUR_USERNAME --format svg --theme dark
# Get actionable profile advice
devcard advise YOUR_USERNAME
# Compare two developers
devcard compare karpathy torvaldsDevCard analyzes public GitHub data to build a structured developer profile with 14+ signals:
| Signal | Source | Example |
|---|---|---|
| Identity | User profile | Name, bio, location, followers |
| Languages | Repo language stats | Python 72% (logic), CSS 8% (presentation) |
| Tech Stack | Dependency files | PyTorch, FastAPI, PostgreSQL |
| Activity | Events + search API | Active, 679 commits/year, consistency 75/100 |
| Top Projects | Repos ranked by impact | Signature project with narrative |
| Collaboration | PRs, issues, orgs | Maintainer style, external contributions |
| Quality | Repo file structure | CI 80%, tests 60% + actionable recommendations |
| Expertise | Topics + stack + stars | Machine Learning (0.95, advanced) |
| Notable Contributions | PRs to popular repos | Merged PRs in keras, flask, kubernetes |
| Coding Habits | Commit patches | Spaces indentation, 39 chars/line avg |
| Commit Quality | Commit messages | Avg 160 chars, 20% multi-line |
| README Depth | README content analysis | 748 words avg, 78% have code blocks |
| Code Reviews | PR review events | 5 reviews given, 3 approved |
| Responsiveness | Comment events | Issue comments, PR discussion activity |
All signals are extracted from public GitHub API data.
Generate a DevCard for any public GitHub user.
devcard generate torvalds # Terminal output (default)
devcard generate torvalds --format svg --theme dark # Dark SVG card
devcard generate torvalds --format json -o out.json # Save JSON to file
devcard generate torvalds --token ghp_xxx # Explicit token
devcard generate torvalds --enrich # AI-powered insights
devcard generate torvalds --no-cache # Fresh data
devcard generate torvalds --verbose # Debug loggingGet actionable profile advice with scores, praise, and critiques.
devcard advise chiruu12 # Rules-based advice (no LLM needed)
devcard advise chiruu12 --enrich # Add LLM-generated summary
devcard advise chiruu12 --format json # Machine-readable advice
devcard advise chiruu12 --format markdown # GFM outputOutputs:
- Human Visibility Score (0-100) -- how visible to recruiters
- Agent Readiness Score (0-100) -- how readable by AI tools
- Verdicts -- praise (what's good), critiques (what to fix), suggestions (nice-to-have)
- Each critique includes a specific action to fix it
Compare two developers side by side.
devcard compare karpathy torvalds # Terminal comparison
devcard compare karpathy torvalds --format json # JSON with both cardsValidate a devcard.json against the schema.
devcard validate my-devcard.jsonGenerate a DevCard for your own GitHub account (detected via gh CLI).
devcard me --format svg --theme neondevcard generate <username> --format terminal # Rich terminal panels (default)
devcard generate <username> --format json # Structured devcard.json
devcard generate <username> --format yaml # Human-friendly YAML
devcard generate <username> --format svg # Embeddable SVG card
devcard generate <username> --format markdown # GitHub Flavored Markdown
devcard generate <username> --format agent # Machine-readable key:value
devcard generate <username> --format llms-txt # llms.txt spec for AI agents
devcard generate <username> --format all # Everything at onceDevCard outputs are designed for machine consumption, not just human eyes.
devcard generate torvalds --format agent # Key:value pairs for AI tools
devcard generate torvalds --format llms-txt # llms.txt protocol spec
devcard generate torvalds --format json # Structured devcard.jsonDevCard ships as an MCP server, so AI agents like Claude can query developer profiles directly:
# An AI agent calls:
get_devcard("torvalds") # Returns typed fields for every signal
compare("karpathy", "torvalds") # Side-by-side structured comparisonSee devcard-mcp/README.md for MCP setup.
With a Fireworks API key, --enrich adds LLM-powered analysis:
- Developer archetype -- creative label like "ML Craftsman" or "Full-Stack Polyglot"
- AI summary -- 1-2 sentence narrative citing your actual projects
- Evidence-backed strengths -- "Strong ML foundations -- uses PyTorch across 3 projects"
- Actionable suggestions -- "Add CI to FungiClassifier -- it's your top project"
- Smart project ranking -- by significance, not just star count
export FIREWORKS_API_KEY=your-key
pip install devcard[enrich] # Install AI dependencies
devcard generate karpathy --enrich # Generate with AI insights
devcard advise chiruu12 --enrich # Advice with AI summary| Default | Dark |
| Minimal | Neon |
| Terminal Green | |
-
Generate your SVG:
devcard generate YOUR_USERNAME --format svg --theme dark -o devcard.svg
-
Add
devcard.svgto your profile repo (YOUR_USERNAME/YOUR_USERNAME) -
Reference it in your
README.md:<p align="center"> <img src="devcard.svg" alt="My DevCard" width="495"> </p>
DevCard ships with 4 skills for Claude Code users. Add them to your Claude Code config to get guided workflows:
| Skill | What It Does |
|---|---|
devcard |
MCP tool orchestration — audit, compare, fix profiles |
devcard-profile-audit |
Full audit + advise + fix cycle |
devcard-generate |
Card generation in all formats |
devcard-agent-ready |
Make your profile AI-agent readable |
Skills are in skills/claude-code/. To use them, point your Claude Code skill path to this directory.
Every DevCard produces a devcard.json -- a structured, agent-readable profile:
{
"$schema": "https://devcard.dev/schema/v1",
"version": "1.0",
"identity": {
"username": "karpathy",
"name": "Andrej Karpathy",
"bio": "I like to train neural nets",
"followers": 95000
},
"languages": [
{ "name": "Python", "percentage": 72.3, "category": "logic" }
],
"expertise": {
"domains": [
{ "name": "Machine Learning", "confidence": 1.0, "skill_level": "expert" }
],
"profile_type": "ml"
}
}The full schema is at schema/devcard.v1.schema.json.
CLI (typer) --> Pipeline
|-- GitHub Client (async httpx, cached, rate-limited)
|-- Extractors (14 parallel: identity, languages, stack, activity,
| projects, collaboration, quality, expertise, notable,
| habits, reviews, lines, commit_quality, readme_depth,
| responsiveness)
|-- Analyzers (developer type, project classifier, contribution style, scoring)
|-- Advisor (YAML rules engine + optional LLM)
|-- Enrichment (optional LLM via Fireworks AI)
|-- Renderers (terminal, SVG, markdown, JSON, YAML, agent, llms.txt)
git clone https://github.com/chiruu12/devcard.git
cd devcard
uv sync --dev # Install dependencies
uv run pytest # Run tests (361 tests)
uv run ruff check src/ tests/ # Lint
devcard generate chiruu12 # Test locally
devcard advise chiruu12 # Test advisorSee CONTRIBUTING.md for contribution guidelines.
- Terminal, SVG, JSON, YAML, Markdown output
- Compare command
- AI enrichment via Fireworks (
--enrich) - Profile advisor with YAML rules engine (
devcard advise) - 14 extraction signals (identity through responsiveness)
- Dual scoring (human visibility + agent readiness)
- Notable contributions detection
- MCP server (10 tools)
- Claude Code skills (4 skills)
- PyPI publishing --
pip install devcard - GitHub Action -- auto-update your DevCard SVG on push
- PNG export -- for social sharing
- Web app -- connect GitHub, generate your card, share a link