AutoGit

Run autogit in any project folder. It reads your code, writes your docs, commits, pushes to GitHub, updates your resume, and generates your LinkedIn post — in one command.

npm install -g autogit-tool
autogit

What it does

When you run autogit inside a project directory it:

1. Scans the codebase and detects languages, frameworks, databases, CI/CD, deployment config
2. Generates or improves your README
3. Creates PROJECT_SUMMARY.md, ARCHITECTURE.md, and CONTRIBUTING.md if missing
4. Generates a commit message using AI (Conventional Commits format)
5. Stages, commits, and pushes to GitHub — creating the repository if it doesn't exist
6. Updates your LaTeX resume with an AI-written project entry
7. Generates a LinkedIn post, X (Twitter) post, DEV.to draft, and resume bullet
8. Saves a share package with a sourced LinkedIn post, text drafts, evidence, and three PNG cards; offers a Brag video when its optional setup is installed

The share package is saved under ~/.autogit/social/<project>/. LinkedIn publishing and image upload still require your review. Destructive or public-facing Git actions require confirmation; pass --yes to skip prompts. AutoGit can be run from a wrapper containing one project, but refuses to treat your user home directory as a project.

The main run creates three evidence cards automatically, including for CLI, API, and library projects. A real screenshot is optional and only applies when a web app is currently reachable at an HTTP(S) URL; leave the URL blank to skip it. Conceptual artwork is available with autogit --promo-image when an image provider is configured.

Installation

npm install -g autogit-tool

npm page: https://www.npmjs.com/package/autogit-tool

Requirements: Node.js 18+, Git

On the first interactive autogit run, setup asks for Default (preselected) or Customize, a text AI provider, a separate image provider, and whether to include Brag (Yes preselected). Codex, Claude Code, and Antigravity use their own browser sign-in and subscription quota for text generation. Images can use Antigravity's signed-in quota, OpenAI, Gemini, xAI, Together AI, or a custom OpenAI-compatible image endpoint. Rerun the wizard with autogit setup. npm does not reliably expose interactive package-install prompts, so the questions appear on first launch, not while npm install is running. Non-interactive runs and --yes skip the wizard.

The optional Brag step installs the selected agent and its skill globally. Run autogit brag from a project folder to start the video workflow. AutoGit shows animated setup and rendering progress, then prints the exact video path and a clickable link to its output folder. Codex, Claude Code, and Antigravity use their own cached OAuth sessions; AutoGit never passes them an API key. OpenCode receives only the selected provider credential for that one process and does not store it in OpenCode settings. During opted-in setup, AutoGit installs Node 22 and FFmpeg into its user-data folder if needed; Brag uses npx hyperframes to fetch its renderer on first use. The managed FFmpeg binary is a third-party GPL-licensed component. The image-card showcase works without those video tools. Azure OpenAI is not yet mapped to the OpenCode runner.

Quick start

# 1. Go to any project folder
cd my-project

# 2. First run opens the Default/Customize and AI setup wizard
autogit setup

# 3. Set up your GitHub token (first time only)
autogit login

# 4. Run
autogit

AI Providers

AutoGit supports API providers plus three OAuth-backed coding-agent providers. Configure via autogit config or environment variables.

| Provider | Env Variable | Default Model | Free Tier | |---|---|---|---| | Groq | GROQ_API_KEY | llama-3.3-70b-versatile | ✔ Yes | | OpenAI | OPENAI_API_KEY | gpt-4o-mini | ✘ No | | Anthropic | ANTHROPIC_API_KEY | claude-sonnet-4-20250514 | ✘ No | | Google Gemini | GEMINI_API_KEY | gemini-1.5-flash | ✔ Yes | | Mistral | MISTRAL_API_KEY | mistral-large-latest | ✔ Limited | | DeepSeek | DEEPSEEK_API_KEY | deepseek-chat | ✔ Limited | | Perplexity | PERPLEXITY_API_KEY | llama-3.1-sonar-large-128k-online | ✘ No | | Together AI | TOGETHER_API_KEY | meta-llama/Llama-3.3-70B-Instruct-Turbo | ✔ Limited | | Cohere | COHERE_API_KEY | command-r-plus-08-2024 | ✔ Limited | | xAI (Grok) | XAI_API_KEY | grok-3-fast-beta | ✘ No | | OpenRouter | OPENROUTER_API_KEY | anthropic/claude-sonnet-4-20250514 | ✔ Free models | | Ollama | OLLAMA_ENDPOINT | llama3.1 | ✔ Local | | Azure OpenAI | AZURE_OPENAI_KEY | your deployment | — | | NVIDIA NIM | NVIDIA_API_KEY | meta/llama-3.3-70b-instruct | ✔ Free credits | | Cerebras | CEREBRAS_API_KEY | gpt-oss-120b | — | | DeepInfra | DEEPINFRA_API_KEY | deepseek-ai/DeepSeek-V3.2 | — | | Hugging Face Inference | HUGGINGFACE_API_KEY or HF_TOKEN | openai/gpt-oss-120b:fastest | ✔ Limited | | Fireworks AI | FIREWORKS_API_KEY | accounts/fireworks/models/llama-v3p1-8b-instruct | — | | Codex | ChatGPT browser sign-in | ChatGPT subscription | subscription quota | | Claude Code | Claude browser sign-in | Claude subscription | subscription quota | | Antigravity | Google browser sign-in | Google account subscription | account quota | | Custom OpenAI-compatible | CUSTOM_API_KEY | your model | — |

Recommended for free usage: Groq — fastest free API, no credit card required. NVIDIA NIM also provides free credits on signup.

Configure interactively

autogit config
# → AI Provider & Keys → select provider → paste key

Configure via environment variable

# PowerShell
$env:AUTOGIT_AI_PROVIDER = "groq"
$env:GROQ_API_KEY = "gsk_your_key_here"

# Bash / Zsh
export AUTOGIT_AI_PROVIDER=groq
export GROQ_API_KEY=gsk_your_key_here

# Windows CMD
set AUTOGIT_AI_PROVIDER=groq
set GROQ_API_KEY=gsk_your_key_here

Verify your key

autogit config --test

Custom endpoint

Choose custom in autogit config to use any service that implements the OpenAI chat-completions API. Enter its base endpoint, model name, and API key. The key is optional for local servers such as LM Studio, LocalAI, vLLM, and llama.cpp. The same endpoint, model, and key are passed to OpenCode for the optional Brag workflow; credentials are not saved in OpenCode.

Image providers

Image generation is configured separately, so the text and image providers can differ. Antigravity can create artwork through its Google sign-in and included quota without an image API key.

| Provider | Credential | Default image model | |---|---|---| | Antigravity | Google sign-in | Built-in image generation | | OpenAI | OPENAI_API_KEY | gpt-image-1.5 | | Google Gemini | GEMINI_API_KEY | gemini-3.1-flash-image | | xAI | XAI_API_KEY | grok-imagine-image-2.0 | | Together AI | TOGETHER_API_KEY | black-forest-labs/FLUX.2-dev | | Custom OpenAI-compatible | AUTOGIT_IMAGE_API_KEY | user supplied |

Run autogit setup or choose Image Provider in autogit config. Environment overrides are AUTOGIT_IMAGE_PROVIDER, AUTOGIT_IMAGE_MODEL, AUTOGIT_IMAGE_ENDPOINT, and AUTOGIT_IMAGE_API_KEY.

GitHub Authentication

autogit login

The login wizard will:

• Auto-import your token if the GitHub CLI is installed and authenticated
• Otherwise prompt you to paste a Personal Access Token

Create a token: https://github.com/settings/tokens/new Required scopes: repo, read:user

You can also set it via environment variable:

export GITHUB_TOKEN=github_pat_your_token_here

Troubleshooting bad credentials:

autogit login --check   # shows which token source is active and verifies it

If GITHUB_TOKEN is set in your environment and invalid, clear it:

# PowerShell — current session only
Remove-Item Env:GITHUB_TOKEN

# PowerShell — permanent (removes from Windows user environment)
[System.Environment]::SetEnvironmentVariable("GITHUB_TOKEN", $null, "User")

Resume Auto-Update

AutoGit can automatically add a project entry to your LaTeX resume every time you run autogit.

Setup (one time)

autogit resume --setup

Point it at your .tex resume file. AutoGit will:

• Detect your \section{Projects} block
• Insert an AI-generated \resumeProjectHeading entry at the top
• Create a timestamped .backup-<timestamp>.tex before writing
• Offer to copy the updated file to ~/Documents/resume/

Usage

autogit resume               # update resume for current project
autogit resume --show        # check configured path
autogit resume --setup       # reconfigure
autogit resume --file resume.tex         # use an existing LaTeX file
autogit resume --from-pdf resume.pdf     # convert a text-based PDF to LaTeX

PDF conversion extracts text locally, sends the extracted resume content to your configured AI provider, and writes a .tex file beside the PDF when that location is writable. Scanned, image-only PDFs require OCR first. Review the generated LaTeX before using it.

The resume step runs automatically in the main pipeline. Skip it with:

autogit --skip-resume

All Commands

| Command | Description | |---|---| | autogit | Full pipeline: scan → docs → resume → commit → push → share package, with optional Brag video | | autogit init | Initialize Git and generate .gitignore | | autogit docs | Generate PROJECT_SUMMARY.md, ARCHITECTURE.md, CONTRIBUTING.md | | autogit readme | Generate or update README only | | autogit publish | Commit and push to GitHub | | autogit github | Create GitHub repository | | autogit linkedin | Generate all social content (short/medium/long LinkedIn, Twitter, DEV.to, resume bullet) | | autogit setup | Configure defaults, text and image providers, and optional Brag | | autogit brag | Run Brag video creation with the matched coding agent | | autogit showcase | Create a sourced project post and three LinkedIn-ready PNG cards | | autogit release | Create a GitHub release with tag and notes | | autogit resume | Update LaTeX resume with current project | | autogit analyze | Project score card: documentation, code quality, maintainability | | autogit doctor | Check that Git, Node.js, GitHub token, and AI keys are all working | | autogit config | Interactive configuration wizard | | autogit login | Authenticate with GitHub |

Publishing Safely

Before committing, AutoGit shows files already staged, files it plans to stage, the current branch, and the origin remote. Partially staged files keep their staged version; their remaining edits stay unstaged. Obvious credential files such as .env and private keys stop the commit. A failed push returns an error, while the local commit remains available to retry.

If origin exists, AutoGit pushes there without changing its URL or requiring a GitHub API token. If there is no remote, AutoGit can create a new repository; new repositories are private by default unless --public or a saved public visibility default is specified. If a same-named repository already exists, AutoGit asks you to set origin yourself so it cannot choose the wrong destination. --yes skips prompts but does not bypass the credential-file check.

LinkedIn Images

For any project, including a CLI, API, library, or data project, run autogit showcase. It detects the project type, uses traceable project facts, and saves a post, three PNG cards, editable SVG versions, and an evidence.json file under ~/.autogit/social/<project>/showcase-<timestamp>/. Review the draft before sharing. It never runs project commands or invents terminal output, performance results, or a UI screenshot. The PNG renderer is included in AutoGit's npm dependencies; no coding agent, browser, FFmpeg, or Hyperframes installation is required. Showcase generation requires the configured AI provider. Videos are not generated by this command.

Run autogit linkedin and choose a real screenshot, conceptual promotional artwork, or both. Images are saved under AutoGit's user data directory and must be uploaded in LinkedIn's composer; opening the share dialog does not attach local image files.

autogit linkedin --screenshot-url http://localhost:3000
autogit linkedin --promo-image
autogit linkedin --showcase-cards
autogit linkedin --screenshot-url http://localhost:3000 --promo-image "minimal product illustration"

Screenshots capture a running web app using Chrome, Edge, or Chromium installed on the computer. Promotional artwork uses the separately configured Antigravity, OpenAI, Gemini, xAI, Together AI, or custom image provider. Antigravity uses Google sign-in and its included quota instead of an image API key. Artwork is conceptual and should not be presented as a screenshot of the app.

All Flags

| Flag | Description | |---|---| | --yes | Skip all confirmation prompts | | --dry-run | Preview all changes without writing anything | | --verbose | Show detailed output | | --quiet | Suppress non-essential output | | --private | Create private GitHub repository | | --public | Create public GitHub repository | | --skip-readme | Skip README generation | | --skip-github | Skip GitHub operations | | --skip-linkedin | Skip social content generation | | --skip-resume | Skip resume auto-update | | --regenerate | Force-regenerate existing documentation files |

Configuration

Settings are stored in ~/.autogit/config.json.

autogit config             # interactive wizard
autogit config --list      # view current config (secrets masked)
autogit config --debug     # show which env vars and config keys are active
autogit config --test      # verify active AI provider key with a live call
autogit config --set ai.provider=groq   # set a single value
autogit config --get ai.provider        # read a single value

Full config structure

{
  "github": {
    "token": "github_pat_..."
  },
  "ai": {
    "provider": "groq",
    "model": "",
    "groqKey": "gsk_..."
  },
  "defaults": {
    "visibility": "public",
    "branch": "main",
    "license": "MIT",
    "commitStyle": "conventional",
    "autoConfirm": false,
    "linkedinStyle": "professional"
  },
  "resume": {
    "path": "/Users/you/Documents/resume/resume.tex",
    "ownerName": "Your Name",
    "ownerEmail": "[email protected]",
    "enabled": true
  }
}

Project Detection

AutoGit auto-detects the following without any configuration:

Languages: TypeScript, JavaScript, Python, Rust, Go, Java, Kotlin, C#, PHP, Ruby, Swift, C/C++, Dart, Elixir, Scala

Frameworks: React, Next.js, Vue, Nuxt, Angular, Svelte, Express, Fastify, NestJS, Django, Flask, FastAPI, Spring Boot, Actix, Axum, Gin, Echo and more

Databases: PostgreSQL, MySQL, MongoDB, Redis, SQLite, DynamoDB, Supabase, Firebase

Deployment: Docker, Vercel, Netlify, Fly.io, Render, Heroku, AWS SAM, Terraform, Serverless Framework

CI/CD: GitHub Actions, GitLab CI, CircleCI, Jenkins, Travis CI, Azure Pipelines

Ignored directories: node_modules, .git, dist, build, coverage, .next, vendor, target, __pycache__ and more

Dry Run

Preview everything before it runs:

autogit --dry-run

Shows what would be changed without touching any file, making any commit, or calling GitHub.

Project-aware README generation

With an AI provider configured, autogit readme inspects the current project in several steps. It can search the project file list and read relevant source and configuration files before drafting. AutoGit shows the proposed README or a diff and asks before writing. No separate coding-agent program is required; the feature uses AutoGit's existing AI provider configuration.

Sensitive files such as .env, private keys, and lockfiles are excluded from this inspection. If AI generation fails or no provider is configured, the command exits with an error and does not write a README.

Contributing

Pull requests are welcome.

git clone https://github.com/vardhchhajer/autogit-tool.git
cd autogit-tool
npm install
npm run build        # compile TypeScript → dist/
npm link             # install autogit globally from this local folder

Development scripts:

| Script | What it does | |---|---| | npm run build | Compile TypeScript to dist/ (required before running) | | npm test | Build and run the README agent tests | | npm run dev | Watch mode — recompiles automatically on every file save | | npm start | Run the CLI directly via node dist/cli.js | | npm run lint | Run ESLint across src/ | | npm link | Register autogit as a global command pointing at this local build | | npm publish --access public | Publish a new version to npm |

Project structure:

src/
├── cli.ts                    # Entry point — all commands registered here
├── commands/                 # One file per subcommand
├── config/manager.ts         # Config file + env var resolution
├── scanner/                  # File scanner + project analyzer
├── ai/                       # Provider abstraction + prompts
├── services/                 # readme, docs, git, github, resume, social, insights
├── pipeline/main-pipeline.ts # Orchestrates the full autogit flow
└── utils/                    # Logger, platform helpers

License

MIT © Vardh Chhajer

npm: https://www.npmjs.com/package/autogit-tool GitHub: https://github.com/vardhchhajer/autogit-tool