Pack your entire codebase into a single AI-friendly file.
Feed your repository to Claude, ChatGPT, Gemini, and other LLMs -- with built-in security scanning, cost estimation, and prompt templates.
LLMs work best when they can see your entire codebase at once. But copying and pasting files manually is tedious, error-prone, and dangerous -- you might accidentally send API keys or credentials to a third-party model.
repogrok solves this. One command turns your repository into a single, structured file that is optimized for AI consumption. It scans for secrets before they leave your machine, counts tokens so you know the cost before you paste, and includes prompt templates so the AI knows exactly what to do.
npx repogrok
That is it. Zero config. Works out of the box.
npx repogrok# npm
npm install -g repogrok
# yarn
yarn global add repogrok
# homebrew (coming soon)
brew install repogrok# Pack the current directory into repogrok-output.xml
repogrok
# Pack with a specific prompt template
repogrok --prompt code-review
# Pack only changed files for a PR review
repogrok --diff
# See token counts and estimated LLM costs
repogrok --tokens
# Copy output directly to clipboard
repogrok --copy
# Pack a remote GitHub repository
repogrok --remote facebook/reactrepogrok traverses your repository, respects your .gitignore and .repogrokignore rules, and produces a single file with every source file neatly structured. The output includes a file tree, metadata, and the full contents of each file with clear delimiters.
Choose the format that works best for your workflow:
| Format | Flag | Best For |
|---|---|---|
| XML | --format xml |
Claude, structured parsing |
| Markdown | --format markdown |
ChatGPT, human readability |
| JSON | --format json |
Programmatic access, pipelines |
| Plain | --format plain |
Simple text, universal compat |
XML is the default because it provides the clearest structure for most LLMs.
Before your code reaches any LLM, repogrok scans every file for:
- API keys and secret keys
- Passwords and credentials
- Authentication tokens
- Private keys and certificates
- Connection strings with embedded credentials
If a secret is detected, repogrok warns you and excludes the offending content. Your credentials never leave your machine.
Know exactly how much an AI query will cost before you send it:
repogrok --tokensrepogrok -- Token & Cost Report
--------------------------------------------
Total files: 142
Total tokens: 87,341
Estimated cost per query:
Claude Opus 4 $1.31 (input) + ~$0.40 (output)
Claude Sonnet 4 $0.26 (input) + ~$0.12 (output)
GPT-4o $0.22 (input) + ~$0.11 (output)
Gemini 2.5 Pro $0.11 (input) + ~$0.07 (output)
--------------------------------------------
No more guessing. No more surprise bills.
Skip the prompt engineering. repogrok ships with seven battle-tested templates that tell the LLM exactly what to do with your codebase:
repogrok --prompt <template>| Template | What It Does |
|---|---|
code-review |
Comprehensive code review with severity-ranked findings |
refactor |
Identify refactoring opportunities with concrete suggestions |
document |
Generate documentation for undocumented code |
security |
Deep security audit focused on vulnerabilities and exploits |
onboard |
Produce an onboarding guide for new developers |
test |
Generate test cases and identify untested code paths |
migrate |
Analyze migration paths for frameworks, languages, or versions |
Each template is carefully crafted to produce actionable, structured output from the LLM.
Working on a pull request? Pack only the files that changed:
repogrok --diffThis compares your working tree against the base branch and packs only modified, added, or renamed files. Perfect for focused PR reviews without burning tokens on unchanged code.
Analyze any public GitHub repository without cloning it first:
repogrok --remote owner/reporepogrok clones the repository to a temporary directory, packs it, and cleans up after itself.
Include recent commit history for additional context:
repogrok --include-logsThe AI can use commit messages to understand why code was written, not just what it does.
Send the output straight to your clipboard, ready to paste into any AI chat interface:
repogrok --copyrepogrok automatically respects your .gitignore file. For additional exclusions specific to AI packing, create a .repogrokignore file in your project root:
# .repogrokignore
node_modules/
dist/
*.min.js
coverage/
*.lock
The default XML output is structured for maximum LLM comprehension:
<repogrok>
<meta>
<repository>my-project</repository>
<generated>2026-02-16T12:00:00Z</generated>
<file_count>142</file_count>
<total_tokens>87341</total_tokens>
</meta>
<file_tree>
src/
index.ts
utils/
helpers.ts
validate.ts
package.json
tsconfig.json
</file_tree>
<files>
<file path="src/index.ts" language="typescript" tokens="342">
import { helpers } from './utils/helpers';
import { validate } from './utils/validate';
export function main() {
const config = helpers.loadConfig();
validate(config);
return config;
}
</file>
<file path="src/utils/helpers.ts" language="typescript" tokens="198">
export const helpers = {
loadConfig() {
// ...
}
};
</file>
</files>
</repogrok>Usage: repogrok [directory] [options]
Arguments:
directory Target directory (default: current directory)
Output Options:
-o, --output <file> Output file path (default: repogrok-output.xml)
-f, --format <type> Output format: xml, markdown, json, plain (default: xml)
--copy Copy output to clipboard
--tokens Display token counts and cost estimates
Content Options:
--prompt <template> Apply a prompt template: code-review, refactor,
document, security, onboard, test, migrate
--diff Pack only changed files (compared to base branch)
--remote <owner/repo> Pack a remote GitHub repository
--include-logs Include recent git commit history
Filter Options:
--include <globs> Include only files matching these patterns
--exclude <globs> Exclude files matching these patterns
--no-gitignore Do not respect .gitignore rules
--max-file-size <bytes> Skip files larger than this size
Security Options:
--no-security-scan Disable secret detection (not recommended)
General:
-v, --version Show version number
-h, --help Show help
For project-specific defaults, create a repogrok.config.json in your project root:
{
"output": "repogrok-output.xml",
"format": "xml",
"include": ["src/**", "lib/**", "package.json", "tsconfig.json"],
"exclude": ["**/*.test.ts", "**/*.spec.ts", "dist/**"],
"maxFileSize": 524288,
"securityScan": true,
"includeGitLogs": false,
"prompt": null
}Command-line flags always override configuration file values.
| Feature | repogrok | Repomix |
|---|---|---|
| Pack codebase to one file | Yes | Yes |
| XML output | Yes | Yes |
| Markdown output | Yes | Yes |
| JSON output | Yes | Yes |
| Plain text output | Yes | Yes |
| Security scanning | Yes | Yes |
| Token counting | Yes | Yes |
| LLM cost estimation | Yes | No |
| Built-in prompt templates | Yes | No |
| Diff-only mode | Yes | No |
| Remote repo support | Yes | Yes |
| Git log inclusion | Yes | No |
| Clipboard copy | Yes | Yes |
| .gitignore support | Yes | Yes |
| Custom ignore file | Yes | Yes |
| Zero config | Yes | Yes |
Code review -- Pack your PR changes with --diff --prompt code-review and get a thorough, structured review from any LLM in seconds.
Onboarding -- New to a codebase? Run repogrok --prompt onboard and get a complete walkthrough of the architecture, key files, and conventions.
Security audit -- Run repogrok --prompt security to have an LLM scan your entire codebase for vulnerabilities, injection risks, and insecure patterns.
Documentation -- Generate missing documentation across your project with repogrok --prompt document.
Migration planning -- Evaluating a framework upgrade? Run repogrok --prompt migrate to get a detailed migration plan with risk assessment.
Test generation -- Identify untested code paths and generate test scaffolding with repogrok --prompt test.
Refactoring -- Find code smells, duplication, and improvement opportunities with repogrok --prompt refactor.
Contributions are welcome. Whether it is a bug fix, a new prompt template, an output format improvement, or a documentation update -- we appreciate it.
- Fork the repository at github.com/glue-tools-ai/repogrok
- Create your feature branch:
git checkout -b feature/my-feature - Make your changes
- Run the test suite:
npm test - Commit your changes:
git commit -m "Add my feature" - Push to the branch:
git push origin feature/my-feature - Open a pull request
Please open an issue first for major changes so we can discuss the approach.
MIT -- see LICENSE for details.
repogrok is built and maintained by Glue. We build developer tools that make working with AI practical, safe, and efficient.
getglueapp.com | GitHub | npm
