Potencia tu agente
El comando propio del publicador. No es un endpoint de PlugMCP.
Host a copy with our keys and logs.
npx -y [email protected]Claude Code no tiene enlace de instalación para un comando local. Copia el comando de abajo y ejecútalo en una terminal.
claude mcp add bagwan-kai-bandey -- npx -y bagwan-kai-bandey@0.3.7Ejecútalo en una terminal.
{ "mcpServers": { "bagwan-kai-bandey": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
¿URL remota o un comando en tu máquina? En qué se diferencian las dos instalaciones
Ejecutar en PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
AbrirChoose how you want to run BKB:
# Option A: Global CLI (Recommended for fast daily workflows & clean 'bkb' commands)
npm install -g bagwan-kai-bandey
bkb doctor
# Option B: Zero-Install via npx (Instant execution without local installation)
npx bagwan-kai-bandey doctor
# Option C: From Source (For local development & contributors)
git clone https://github.com/saqib40/bagwan-kai-bandey.git
cd bagwan-kai-bandey && npm install
./bin/bkb.js doctor
CLI Command Syntax: Throughout this documentation, commands use the short
bkbsyntax (e.g.bkb init,bkb cockpit,bkb status). If you prefer zero-installnpx, simply prependnpx bagwan-kai-bandeyinstead (e.g.npx bagwan-kai-bandey init). Both forms are 100% interchangeable.
docker run -p 6379:6379 -p 3000:3000 -d --name falkordb-bkb-engine falkordb/falkordb:latest
(Note: BKB will also automatically launch this container if Docker is running).Want to see BKB in action without touching your own code? BKB includes a built-in sandbox (./demo-repo featuring an Express.js app with circular dependencies, authentication middleware, and integration tests).
# 1. Run environment diagnostics (verifies Node, Docker, and FalkorDB)
bkb doctor
# 2. Ingest the built-in demo sandbox into FalkorDB
bkb init --repo ./demo-repo --pack express-to-fastify
# 3. Launch the live visual flight deck in your browser
bkb cockpit
Open http://localhost:4200 in your browser. You will see the live dependency graph, PageRank risk scores, Tarjan SCC cycle clusters, and invariant verification gates standing by.
Choose your preferred way of working:
┌────────────────────────────────────────────────────────────────────────┐
│ HOW DO YOU WANT TO MIGRATE? │
├───────────────────────────────────┬────────────────────────────────────┤
│ Workflow A: AI Agent via MCP │ Workflow B: Autonomous CLI │
│ (Cursor / Claude / OpenCode) │ (Terminal / CI Execution) │
│ • Agent calls graph tools │ • Single command execution │
│ • Generates idiomatic code │ • Uses Mock, Ollama, or Cloud LLM │
│ • Commits milestones to Git │ • Automated rollback safety net │
└───────────────────────────────────┴────────────────────────────────────┘
BKB exposes a standardized Model Context Protocol (MCP) server that turns Cursor, Claude Desktop, VS Code, Windsurf, or OpenCode into a graph-governed migration engineer.
Register the BKB MCP server in your editor with a single command:
# Target your specific editor:
bkb mcp install cursor # Cursor (.cursor/mcp.json)
bkb mcp install claude # Claude Desktop (claude_desktop_config.json)
bkb mcp install vscode # VS Code (.vscode/mcp.json)
bkb mcp install windsurf # Windsurf (.codeium/windsurf/mcp_config.json)
bkb mcp install opencode # OpenCode CLI / opencode.jsonc
# Or install for all detected editors at once:
bkb mcp install all
# Ingest your codebase and capture pre-migration invariant baseline
bkb init --repo /path/to/your-repo --pack express-to-fastify
# In another terminal window, launch the real-time visual cockpit
bkb cockpit --repo /path/to/your-repo
Paste one of these prompts into your agent's chat interface:
You are an autonomous codebase migration engineer governed by Bagwan Kai Bandey (BKB) MCP tools.
Your objective: Safely and incrementally migrate this codebase using the active migration pack, preserving all compiler gates, invariant snapshots, and backward compatibility.
Protocol:
1. CHECK STATUS:
Start by calling `bkb_get_migration_status` to inspect overall progress, the active phase (SHIM, MIGRATE, or DE_SHIM), and unblocked leaves ready for migration.
2. CLAIM & INSPECT BATCH:
Call `advance_migration` with NO arguments (`{}`) to claim the next topological batch.
- The response provides the full disk source, callee symbol signatures, caller snippets, and specific pack guidance.
- Do NOT grep or crawl the workspace; everything needed is in the returned slice.
3. TRANSFORM WITH DUAL COMPATIBILITY:
Write the migrated code adhering strictly to the returned slice instructions:
- In SHIM phase: Preserve existing exports and caller call patterns (using @bkb-shim dual branches or runtime-interop export shapes). Downstream callers must not break!
- In DE_SHIM phase: Strip the legacy shims cleanly once all upstream callers are verified.
4. VERIFY & ADVANCE:
Call `advance_migration` with `edits: [{ path: "<path>", content: "<full new code>" }]` and an `agentSummary`.
- The engine writes the file, runs AST re-parsing, executes compiler typechecks (tsc/mypy), and verifies invariant snapshots.
- If verification passes: The engine commits the node in FalkorDB, updates `.bkb/manifests/<packId>.json`, and returns the next batch.
- If verification fails: Inspect the compiler error returned, fix the types, and re-call `advance_migration`.
5. COMMIT MILESTONES:
Whenever an atomic batch or sprint milestone is verified, call `bkb_commit_milestone` to create an atomic Git commit with structured audit notes so other team members can track progress.
Repeat this loop until `bkb_get_migration_status` confirms 100% of modules are VERIFIED.
"Use the BKB MCP server to migrate this repository. Call
bkb_get_migration_statusto see current progress, then run theadvance_migrationloop to claim, transform, and verify each batch until 100% complete."
bkb run)Execute migrations directly from your shell without an interactive editor:
# 1. Zero-Cost Deterministic Dry Run (Simulates the entire topological loop):
bkb run --repo /path/to/your-repo --pack express-to-fastify --dry-run
# 2. Local Zero-Cost LLM via Ollama (e.g. Qwen2.5-Coder):
bkb run --repo /path/to/your-repo --pack express-to-fastify --provider ollama --model qwen2.5-coder:latest
# 3. OpenCode / DeepSeek (Default: deepseek-v4.1-flash-free):
bkb run --repo /path/to/your-repo --pack cjs-to-esm --provider opencode
# 4. Cloud LLMs (OpenAI GPT-4o / Anthropic Claude):
export OPENAI_API_KEY="sk-..."
bkb run --repo /path/to/your-repo --pack express-to-fastify --provider openai --model gpt-4o
If a transformation ever fails unexpectedly, revert your codebase and graph state instantly:
bkb rollback # Revert to the latest verified checkpoint
bkb rollback 3 # Revert to specific Step 3
bkb rollback --list # View all available step snapshots
Bagwan Kai Bandey includes four production-ready packs spanning TypeScript/JavaScript and Python:
in-file-branch: In-file dual-compatibility branches tagged with @bkb-shim markers, automatically pruned during the DE_SHIM phase once callers have migrated.runtime-interop: Emits pure target syntax while preserving CommonJS export shape (export default + export { X as "module.exports" }) under Node $\ge 22.12$ synchronous require(esm).TypeScriptAdapter: Parses TS, TSX, JS, and JSX; resolves ESM .js $\rightarrow$ .ts mappings, relative .cjs/.mjs paths, and tsconfig.json path aliases.PythonAdapter: Tree-sitter WASM parser for Python (.py); resolves relative (..foo) and dotted (app.bar) package imports; prunes virtual environments (.venv, venv) and caches; clusters circular dependencies with Tarjan SCC.packs/PACK_SPEC.schema.json) compiled with Ajv, providing instant schema validation via bkb pack validate <path>.Real enterprise migrations span weeks across distributed teams. BKB enables phased migrations with Git-Native Manifests (.bkb/manifests/<packId>.json) — one manifest per migration pack, so concurrent packs never clobber each other:
your-repo/
├── .bkb/
│ ├── manifests/
│ │ └── <packId>.json # Committed directly to Git at each milestone
│ └── checkpoints/ # Local disk checkpoints for instant rollback
.bkb/manifests/<packId>.json. When engineer Bob clones the repository weeks later (with 40% of modules already shimmed), BKB reuses the committed baseline snapshot rather than re-capturing against a half-migrated codebase.bkb init on any developer machine or CI container automatically reads .bkb/manifests/<packId>.json, restores all completed SHIMMED and VERIFIED node states in FalkorDB, and recomputes unblocked leaves in milliseconds.bkb_commit_milestone to atomically stage modified code and .bkb/manifests/<packId>.json, commit with structured audit notes, and embed the new Git commit SHA into the manifest history.bkb status --repo /path/to/your-repo
Outputs an instant ASCII flight deck progress bar ([■■■■░░░░░░░░░░░░░░░░]), module counts, baseline snapshot health, last team milestone, and unblocked leaves ready for the next sprint.bkb cockpit)Launch the dedicated Live Sync Flight Deck & Visualizer:
bkb cockpit --repo /path/to/your-repo
bkb claim, bkb verify, bkb run) or AI agent calls MCP tools, the browser visualizer immediately mirrors the action.Leaf Claim, De-shim Prioritizer, Circular Cycles (SCC), PageRank Risk, Reverse Blast Radius).Stay up to date with the latest migration packs, AST language adapters, and graph engine improvements:
npx (Default):Append @latest to run the newest version automatically:
npx bagwan-kai-bandey@latest doctor
npm install -g bagwan-kai-bandey@latest
bkb doctor
After updating BKB, re-run the 1-click MCP installer so your editors point to the newest tool definitions:
bkb mcp install all
docker pull falkordb/falkordb:latest
For in-depth architectural details, custom framework extensions, and benchmark results:
MIT © saqib40
Bagwan Kai Bandey (BKB): graph-governed codebase migration engine powered by FalkorDB.
{ "mcpServers": { "bagwan-kai-bandey": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}Herramientas
Aún no hay lista de herramientas. Los servidores remotos la reciben tras la primera comprobación de estado.
Preguntas frecuentes
Copia en esta página la configuración stdio para Cursor, Claude Desktop, Claude Code o VS Code. Usa la instalación del publicador, no un proxy de PlugMCP.
Este registro no indica ninguna variable de entorno obligatoria.
Bagwan Kai Bandey se ejecuta localmente por stdio.
Última comprobación: 2026-11-15. Bagwan Kai Bandey (BKB): graph-governed codebase migration engine powered by FalkorDB.