██╗  █████╗  ███╗   ███╗
    ██║ ██╔══██╗ ████╗ ████║
    ██║ ███████║ ██╔████╔██║
██  ██║ ██╔══██║ ██║╚██╔╝██║
╚████╔╝ ██║  ██║ ██║ ╚═╝ ██║
 ╚═══╝  ╚═╝  ╚═╝ ╚═╝     ╚═╝

jam

The developer-first AI CLI. Cross-language code intelligence from your terminal.

Trace call graphs across Java, SQL, Python, and TypeScript. Impact analysis. AI-powered agentic execution. Works with Copilot & Gemini. Windows, macOS, Linux.

Docs · Install · VSCode Extension

---

What Jam Does

Jam isn't a generic AI assistant. It's the senior dev who's seen everything — direct, opinionated, and warm. Every message, error, and prompt speaks with the same voice: concise, specific, developer-aligned.

• 🔍 Call graph tracing — trace any symbol's callers, callees, and upstream chain across languages
• 💥 Impact analysis — "if I change this, what breaks?" with column-level SQL dependency tracking
• 🤖 Agentic execution — jam go (interactive) and jam run (one-shot) decompose tasks into parallel subtasks
• 💬 AI chat & ask — streaming responses, multi-turn sessions, stdin/pipe support
• 🩹 Patch workflow — generate diffs, validate, preview, apply with confirmation
• 📊 Code intelligence — explain files, search code, review diffs, generate Mermaid diagrams
• 🔧 Git toolkit — wtf explains state, undo reverses mistakes, standup shows your work
• ✅ Verification — scan for secrets, lint, type-check before you commit
• 🧰 19 zero-LLM utilities — ports, stats, deps, todo, hash, json, env, and more
• 🔌 Any provider — Copilot, Gemini, Ollama, OpenAI, Anthropic, Groq — or bring your own
• 🏠 Local-first — your code never leaves your machine unless you choose a remote provider
• 🖥️ Cross-platform — Windows (PowerShell, cmd), macOS, Linux
• 🔗 MCP + plugins — connect to Model Context Protocol servers, drop in custom commands

---

Install

# npm
npm install -g @sunilp-org/jam-cli

# Homebrew
brew tap sunilp/tap && brew install jam-cli

# Try without installing
npx @sunilp-org/jam-cli doctor

Jam auto-detects the best available AI provider — no API keys needed if you have Copilot or Gemini:

Priority	Provider	Setup
1	GitHub Copilot	VSCode extension — zero config
2	Gemini Code Assist	VSCode extension — zero config
3	Anthropic	export ANTHROPIC_API_KEY=sk-ant-...
4	OpenAI	export OPENAI_API_KEY=sk-...
5	Ollama (default)	ollama serve + ollama pull llama3.2

jam doctor           # verify everything works
jam models list      # see available models
jam models set gpt-4o  # set your preferred model

---

Cookbook

Ask & Chat

jam ask "explain the builder pattern in Go"

# pipe anything
cat schema.sql | jam ask "what tables have no foreign keys?"
git log --since="1 week" -p | jam ask "summarize this week's changes"

# interactive chat with history
jam chat

Agent Engine

# interactive agent console — reads, writes, runs commands
jam go
jam> add retry logic to the HTTP client with exponential backoff

# one-shot autonomous task
jam run "add input validation to all API endpoints" --yes

# fully autonomous with parallel workers
jam run "refactor auth module into separate files" --auto --workers 4

Code Intelligence

# trace a function's call graph
jam trace createProvider
jam trace updateBalance --impact       # what breaks if this changes?
jam trace handleRequest --mermaid      # output as Mermaid diagram
jam trace PROC_PAYMENT --depth 8       # deeper upstream chain

# explain any file
jam explain src/auth/middleware.ts

# search with AI understanding
jam search "where is the rate limiter configured?"

# generate architecture diagram from code
jam diagram

Git Toolkit

jam git wtf          # "3 files staged, 2 conflicts, 1 stash. Here's what happened..."
jam git undo         # undo last commit, last stash, or last merge
jam git standup      # your commits from the last 3 days
jam git cleanup      # preview and delete merged branches
jam git oops         # fix common mistakes (wrong branch, bad commit message)

Dev Utilities (zero LLM)

jam stats            # LOC, languages, complexity hotspots
jam deps             # import dependency graph
jam todo             # find all TODO/FIXME/HACK comments
jam verify           # pre-commit checks: secrets, lint, types
jam ports            # what's listening on which port
jam env              # environment variable diff between shells
jam hash <file>      # MD5/SHA1/SHA256 of any file
jam json <file>      # validate, format, query JSON
jam recent           # recently modified files
jam convert 5kg lb   # unit conversions
jam http GET /users  # quick HTTP requests
jam pack             # analyze npm/pip/cargo package size

Patch & Review

# AI-powered diff summary
jam diff

# code review with risk assessment
jam review

# generate and apply a patch
jam patch "add error handling to the database module"

# auto-generate commit message matching your project's convention
jam commit

---

VSCode Extension

Install from Marketplace

• All commands in the Command Palette
• @jam chat participant in GitHub Copilot Chat
• TODO tree in the sidebar with click-to-navigate
• Copilot or Gemini auto-detected — zero configuration, no API keys
• Status bar indicator shows connection state
• Keeps jam-cli updated automatically

---

Configuration

jam init              # interactive setup wizard
jam config show       # show resolved config

// .jamrc (per-project)
{
  "defaultProfile": "work",
  "profiles": {
    "work": { "provider": "anthropic", "model": "claude-sonnet-4-20250514" },
    "local": { "provider": "ollama", "model": "llama3.2" }
  }
}

jam ask "hello" --profile work     # use Anthropic
jam ask "hello" --profile local    # use Ollama

Supports HTTP proxy (HTTP_PROXY), custom CA certificates (tlsCaPath), configurable timeouts, MCP servers, and plugin loading. Full configuration docs →

---

Links

• 📖 Documentation
• 🧩 VSCode Extension
• 🍺 Homebrew Tap
• 🐛 Issues

Contributing

See CONTRIBUTING.md. PRs welcome.

License

MIT