Context Bunker Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Context Bunker Mcp (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Stop wasting tokens on orientation. Pre-computed codebase intelligence for AI coding tools. One command. Code never leaves your machine.
AI coding agents spend ~80% of tokens just figuring out where things are — re-reading files, tracing imports, grepping for symbols. Every. Single. Session.
Context Bunker fixes this. It's an MCP server that indexes your codebase using tree-sitter AST parsing and gives your AI tools instant access to structural intelligence — dependency graphs, call trees, dead code, cross-session diffs — all from a local SQLite database.
One call replaces 8-16 grep/read calls. ~90% token savings. Zero cloud. Zero API keys.
Works with Claude Code, Cursor, Windsurf, Copilot, and any MCP-compatible client.
Why not just grep? Grep finds text. Context Bunker understands structure — dependency graphs, cross-session memory, dead code detection, and token-efficient summaries. Things grep structurally can't do. Full comparison.
# Run directly — no install needed
bunx context-bunker-mcp /your/project
# Or install globally
bun install -g context-bunker-mcp
# No Bun? npx / npm work tooClaude Code:
# With a project (auto-indexes on startup)
claude mcp add context-bunker -- bunx context-bunker-mcp /your/project
# Without a project (AI calls set_project dynamically)
claude mcp add context-bunker -- bunx context-bunker-mcp
# No Bun? npx work tooCursor / Windsurf / VS Code — add to settings.json:
{
"mcpServers": {
"context-bunker": {
"command": "bunx",
"args": ["context-bunker-mcp", "/your/project"]
}
}
}| Tool | What it does |
|---|---|
🔥 get_changes_since_last_session | What changed since the AI last looked. No more re-orientation. |
🔥 find_unused_exports | Dead code detection — exported but never imported anywhere. |
🔥 find_unused_code | Find dead internal code — functions, classes, variables, types never referenced. |
🔥 get_file_summary | File overview in ~50 tokens. Scan 10 files for the cost of 1. |
🔥 search_by_pattern | Find structural patterns: HTTP calls, env access, error handlers, async, TODOs. |
🧠 get_smart_context | Full file context in 1 call — imports, exports, dependents, types, tests. |
🧠 get_dependency_graph | "If I change this, what breaks?" — transitive import graph with depth control. |
🧠 find_symbol | AST-aware symbol search by name, kind, and scope. Not text matching. |
🧠 find_references | Where a symbol is used, classified by kind (import, call, type annotation). |
🧠 get_call_graph | What a function calls, recursively, as a tree. |
🧠 get_symbol_source | Extract one definition — not the whole file. 80% token savings. |
🧠 get_project_map | Architecture overview: modules, public APIs, relationships. |
🔧 set_project | Point the index at any project directory on the fly. |
🔧 search_code | Semantic search via local TF-IDF. No API keys needed. |
🔧 reindex | Force re-index of the codebase or a single file. |
🔧 get_status | Index health, file counts, token savings estimate. |
🔥 Unique 🧠 Core Intelligence 🔧 Housekeeping
TypeScript, JavaScript (TSX/JSX/MTS/CTS), Python, Go, Rust, Java, C#
Powered by tree-sitter WASM grammars. Adding a new language = one extractor file in src/languages/.
| Runtime | Bun (Node.js fallback) — bun:sqlite = zero native deps, 4-10x faster startup |
| MCP | @modelcontextprotocol/sdk |
| AST | web-tree-sitter (WASM) — no native bindings, works everywhere |
| Storage | SQLite (WAL mode) — single file, survives restarts |
| Search | TF-IDF (local) — zero API keys, no cloud |
| Watch | chokidar — incremental re-index on file changes |
The index lives in your OS cache directory — not inside the project. No .gitignore needed.
| Platform | Location |
|---|---|
| macOS | ~/Library/Caches/context-bunker/<project>/index.db |
| Linux | ~/.cache/context-bunker/<project>/index.db |
| Windows | %LOCALAPPDATA%\context-bunker\<project>\index.db |
Want it inside the project instead? Use --local or set { "storage": "local" } in .context-bunker.json.
MIT
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.