Stop losing documentation in Notion, Google Docs, and scattered README files.
Unify everything. Navigate with AI.
Open source. Works offline. No account needed.
$ kp init my-project
→ Creating Knowledge Pack: my-project
• Language: English
✓ Knowledge Pack created successfully!
$ kp validate
✓ Pack is valid (8 files checked)Finally, a home for all your project knowledge.
APIs, guides, specs, decisions — all connected in a structured, searchable pack.
Documents link to each other with semantic relationships. Never lose context again.
AI agents navigate your knowledge with full context — no more copying and pasting.
No vendor lock-in. Your docs are plain text files. Works with any editor.
Define your style rules. AI follows YOUR standards when contributing docs.
No cloud required. Everything runs locally. Your knowledge stays with you.
One command. No signup. No config.
curl -fsSL https://kpack.dev/install.sh | bash
iwr -useb https://kpack.dev/install.ps1 | iex
Then run kp init my-project to create your first pack
Everything you need to manage your knowledge.
kp init
Create a new Knowledge Pack
kp serve
Browse your docs in the browser
kp validate
Check structure and links
kp search
Find anything in your pack
kp bundle
Export for AI context windows
kp lint
Check style and writing guidelines
kp transfer
Import docs from external sources
kp profile
Manage writing style profiles
kp describe
AI-optimized pack description
Software, books, companies, research — organize any knowledge.
Whether you're documenting an API, writing a book, or building a company wiki — Knowledge Pack gives you a consistent structure that scales from personal notes to enterprise documentation.
API docs, architecture decisions, guides — all versioned with your code.
Characters, worldbuilding, chapters — connected with relationships.
Processes, onboarding, runbooks — tribal knowledge that never gets lost.
Ideas, research, references — linked with semantic relationships.
my-knowledge-pack/
├── manifest.yaml ← Identity & writing rules
├── README.md ← Entry point
├── docs/
│ ├── concepts/ ← Core ideas
│ ├── guides/ ← Step-by-step
│ ├── api/ ← Reference
│ └── decisions/ ← Why we chose X
└── index/
├── topics.yaml ← What exists
├── glossary.yaml ← Key terms
└── relations.yaml ← How things connectkp context auth --area backend
Filter by area
Proven patterns for effective AI-native documentation.
This is the AI's entry point. Make sure it provides a clear overview and navigation hints for the most important topics. Think of it as the "README for AI" - concise but comprehensive enough to understand your knowledge structure.
The description field in frontmatter appears in bundle summaries.
Write it like an abstract—concise but informative. This is what AI reads first when deciding if a document is relevant.
---
title: JWT Authentication
description: OAuth 2.0 JWT implementation with refresh tokens and role-based access control
topics: [auth, security, backend]
confidence: stable
---
Organization-specific terms belong in index/glossary.yaml.
This prevents AI from misinterpreting your jargon and ensures consistent understanding across all documents.
✓ Good
Define: "PVC", "mesh", "tenant", custom acronyms
✗ Skip
Common terms: "API", "JWT", "Docker"
Mark documents as stable, review, or draft.
AI can then weight responses appropriately—relying more on stable docs and flagging drafts as tentative.
Add kp validate --strict to your CI pipeline to catch structural issues before they reach main.
This ensures your knowledge pack remains AI-readable and prevents broken references.
# .github/workflows/validate.yml
- name: Validate Knowledge Pack
run: kp validate --strict
In index/topics.yaml, define areas like api, frontend, backend to organize topics.
Then use --area flags to filter context by domain.
kp stats --area api
kp context auth --area backend