Why Srcpack

AI assistants are transforming how we work with code, but they hit a wall when your codebase doesn't fit in a single prompt. Srcpack solves this by turning your code into organized, indexed bundles that any AI can understand.

The Problem

When you paste code into ChatGPT, Claude, or Gemini, several things go wrong:

• Context limits — Large codebases exceed token limits, forcing you to cherry-pick files
• Lost structure — Flat file dumps lose the semantic relationships between modules
• No references — AI answers mention "that function" but can't point to exact locations
• Manual work — Every conversation requires re-copying the same files

Teams waste time reformatting code for AI instead of getting answers.

Who Benefits

Team Leads and CTOs

Share your codebase with stakeholders who need precise answers:

• "What's the status of the payment integration?"
• "How complete is the new auth system?"
• "What would it take to add multi-tenancy?"

Upload your bundles once. Anyone with access can query the code without setting up a dev environment or reading source files directly.

Developers

Stop manually copying files into chat windows:

• Code reviews — Upload a bundle, ask "Find potential bugs in the checkout flow"
• Refactoring — "What would break if I renamed UserService to AccountService?"
• Debugging — "Where does this error originate and how does it propagate?"
• Architecture — "How do these modules interact? Draw the dependency graph"

New Team Members

Onboard faster by querying the codebase directly:

• "How does authentication work in this project?"
• "Where are database migrations defined?"
• "What's the pattern for adding new API endpoints?"

TIP

The indexed output means AI answers include exact file and line references, not vague descriptions.

Technical Writers

Generate documentation grounded in actual code:

• "Document the public API of this module"
• "Write a getting-started guide based on the example code"
• "Create a changelog from recent changes"

Use Cases

Project Status and Planning

Bundle your codebase and share it with project managers or executives:

Upload: .srcpack/app.txt

"Analyze this codebase. What features are implemented vs stubbed out?
What areas have the most technical debt? Estimate complexity to add OAuth."

INFO

AI gives answers based on real code, not guesses.

Cross-Team Knowledge Sharing

When another team needs to understand your service:

Upload: .srcpack/api.txt

"We need to integrate with your user service. What endpoints are available?
What authentication does it expect? Show example request/response."

TIP

No meetings required. The code explains itself.

Security Reviews

Have AI audit your code for vulnerabilities:

Upload: .srcpack/auth.txt

"Review this authentication code. Check for:
- SQL injection
- Missing input validation
- Insecure token handling
- OWASP Top 10 issues"

Legacy Code Understanding

Before touching unfamiliar code:

Upload: .srcpack/legacy.txt

"This is legacy code I need to modify. Explain:
- What does this system do?
- What are the main entry points?
- What would break if I change the User class?"

Estimation and Scoping

Get AI help with technical estimates:

Upload: .srcpack/app.txt

"We need to add real-time notifications. Based on this codebase:
- What existing patterns should I follow?
- What modules need changes?
- What's the rough scope of work?"

Why Not Just Copy Files?

You could paste files manually, but srcpack provides:

Manual Copy	Srcpack
Cherry-pick files each time	Define bundles once, reuse forever
Flat text dump	Indexed with file numbers and line ranges
AI says "in that file"	AI says "[3] src/auth.ts:L42"
Includes junk files	Respects .gitignore, skips binaries
Different format each time	Consistent output format
Local only	Optional cloud upload for team sharing

Getting Started

npm

npx srcpack init    # Create config from your project
npx srcpack         # Generate bundles

bun

bunx srcpack init
bunx srcpack

pnpm

pnpm dlx srcpack init
pnpm dlx srcpack

yarn

yarn dlx srcpack init
yarn dlx srcpack

Your bundles are ready in .srcpack/ — upload them to any AI and start asking questions.

See the Getting Started guide for full setup instructions.