Your AI reads files. CKB understands your codebase.
CKB gives Claude Code, Cursor, and other AI tools deep knowledge about your code — impact analysis, ownership, architecture, call graphs, dead code, security — so they stop guessing and start knowing.
How It Works
CKB sits between your AI and your code
Claude Code, Cursor, Windsurf, VS Code, Grok...
Analyzes, indexes, answers
Code, git history, CODEOWNERS, tests...
CKB reads your code and git history, builds a knowledge index, and exposes 90+ query tools to your AI assistant via MCP. Your AI gets precise, structured answers instead of reading raw files.
Let's Clear This Up
CKB is not an AI. It makes your AI smarter.
CKB is NOT
- ✕CKB is an AI coding assistant
- ✕CKB replaces Claude Code or Cursor
- ✕CKB modifies your code
- ✕CKB is a linter or static analyzer
CKB IS
- ✓CKB gives your AI assistant deep code knowledge
- ✓CKB makes Claude Code and Cursor dramatically smarter
- ✓CKB only reads and analyzes — never writes
- ✓CKB understands structure, impact, and ownership
The Difference
Same question. Completely different answer.
Here's what happens when your AI assistant has CKB vs. when it doesn't.
"What breaks if I change this function?"
AI reads nearby files. Misses 80% of dependents. Says "it should be fine."
CKB returns: 12 files affected, 8 direct callers, risk HIGH, 5 tests to re-run.
"Summarize this PR" (600+ files)
AI hits context limit. Reads 20 files. Gives incomplete summary.
CKB summarizes the full PR with impact analysis, risk scores, and affected modules.
"Is this code still used?"
AI greps for the function name. Finds one reference. Says "yes."
CKB traces call graph: zero runtime callers, last touched 2 years ago. Dead code candidate (94% confidence).
"Who should review this?"
AI looks at CODEOWNERS. Suggests Team A.
CKB: Team A owns it on paper, but Team B made 89% of recent commits. Suggests @sarah (92% confidence).
What You Can Do
90+ tools. Here are the ones that change everything.
Every tool works instantly after setup. No configuration needed.
Blast Radius
See exactly what breaks before you merge.
Impact analysis with risk scores, affected files, downstream consumers, and breaking change warnings.
PR Intelligence
Summarize 600-file PRs in seconds.
AI-powered PR summaries with risk assessment. No more scrolling through endless diffs.
Call Graphs & Tracing
Trace any function from entry to exit.
Full call graphs, usage tracing, entrypoint discovery. Know exactly how data flows through your system.
Ownership Intelligence
The right reviewer. Every time.
CODEOWNERS + git blame with time decay. Detect ownership drift. Data-driven reviewer suggestions.
Dead Code Detection
Find what nobody uses anymore.
Static + runtime analysis. Confidence-scored candidates. Clean up safely.
Hotspot Tracking
Your worst files, ranked by data.
Churn analysis with trends. See which files are getting worse over time.
Security Scanning
Find exposed secrets automatically.
26+ patterns for API keys, tokens, passwords. 8-factor risk audit.
Architecture Maps
Understand any codebase in minutes.
Module maps, dependency graphs, coupling analysis. Onboard new devs instantly.
API Change Tracking
Know what changed between versions.
Compare APIs, track contracts, detect breaking changes across your monorepo.
Install. Index. Ask.
# Install
$ npm install -g @tastehub/ckb
# Initialize + index your repo
$ cd your-project && ckb init && ckb index
# Connect to your AI tool
$ ckb setup
Done. Your AI now understands your codebase.Free for personal use. No credit card required.
Works With
Any AI tool that speaks MCP
Plus any custom MCP client, CLI scripts, HTTP API, and CI/CD pipelines.
Language Support
Works with your stack
Full Intelligence
Symbol resolution, call graphs, impact analysis, cross-references.
Navigation
Any language with a language server.
References, definitions, symbol search via LSP.
Git Intelligence
Works for every file. No indexer needed.
Ownership, hotspots, coupling, churn, blame — even for proprietary languages.
Large codebase? Monorepo?
We have a step-by-step guide with 10 testing scenarios for enterprise teams.
Enterprise Testing Guide