Skip to content
BLA (Blind LLM Agents)

BLA (Blind LLM Agents)MCP Server

Released
v3.5.1
MIT

Summary

Defensive MCP server and CLI that hardens your own release artifacts (JS bundles, Electron ASAR, WASM, SQLite, model files) against AI reverse-engineering agents.

Features

  • Hardens JS/TS bundles, Electron ASAR, WASM, SQLite, model files and native binaries
  • bla_harden runs discovery, audit, protection and verification in one call
  • Agent-driven source weaving guided by bla_weave_plan and checked by bla_verify_weave
  • Refuses overwrites without confirmation; --apply and --restore promote or undo a build
  • Bundler plugin for Vite, Rollup, webpack and esbuild

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it.
  3. Restart the agent — the "BLA (Blind LLM Agents)" tools will be available.
{
  "mcpServers": {
    "bla-agents": {
      "command": "npx",
      "args": [
        "-y",
        "bla-agents",
        "mcp"
      ]
    }
  }
}

Description

BLA (Blind LLM Agents) is an open-source tool for software authors who want to make their shipped builds expensive for AI reverse-engineering agents to analyse. It runs as a CLI and as an MCP server (bla) that a coding agent drives to audit, protect and verify a project's build output. The project states it is for defensive use on your own software only.

What it does

BLA discovers build outputs (dist/, build/, .asar, .wasm, .sqlite, .safetensors, native binaries and more) and runs audit → protect → verify. Protection combines format-specific shields, such as decoy SourceMaps, WASM and SQLite decoys and string encryption, with an agent-driven step in which your coding agent rewrites gate logic in the source following BLA's weave plan. Results must keep input/output behaviour identical, and --apply / --restore promote or undo the hardened build without touching source files.

MCP tools

About thirty tools, including bla_harden (the one-call path), bla_discover, bla_audit, bla_protect, bla_verify, bla_lint (packaging-footgun scan, also SARIF from the CLI), bla_weave_plan and bla_verify_weave, bla_encrypt_strings, bla_emit_headers for C, Rust and Go, and bla_doctor. Overwrites require explicit confirmation, every result carries nextSteps, and the bundled skill guide is also served as MCP resources.

Setup

Needs Node.js. npx bla-agents setup registers the server and the blind-llm-agents skill with Claude Code, Cursor, Codex, Windsurf, Gemini CLI, VS Code, OpenCode, GitHub Copilot CLI and Antigravity; the config below is the entry it writes. A Vite, Rollup, webpack and esbuild plugin is included for bundle-time protection. Runs locally, no account. MIT licence.

MCP: Ghidra MCP

by bethington

New

MCP server and Ghidra extension with 219 tools for AI-assisted reverse engineering: decompile, rename, retype and comment, plus P-code emulation, live debugging and Ghidra Server workflows.

MCP: IDA Pro MCP

by mrexodia

New

Open-source MCP server that lets an agent drive IDA Pro for reverse engineering — decompile, trace cross-references, rename, retype, patch and debug — in the GUI or headless through idalib.

MCP: ripwire

by Red Hat

Featured

Red Hat's zero-dependency C++23 code-context engine — ranked call graphs and blast-radius analysis, indexing a repo in under half a second with no server and no database.

MCP: Concord

by Concord AI

New

Local-first MCP server that lets Claude Code, Codex, Cursor, Gemini CLI and other coding agents message each other, claim files and hand off tasks in one repo.

Browse all MCP servers →