Local Rust and Python code scanner providing security audits, health checks, and MCP tools with auditory feedback.
Zero-token static analysis, security vulnerability checks, and MCP server integrations for AI coding.
Vibecode Cleaner Fartrun is a local code safety net and Model Context Protocol (MCP) server designed for developers writing software with AI tools. It scans codebases locally without consuming API tokens or sending your source code to external servers. Use it if you need instant static security checks, debt monitoring, and editor integrations in Cursor, Windsurf, or Claude Code.
The tool combines Rust-compiled analysis modules with Tree-sitter AST parsing to perform 10 security checks and a 9-phase project health audit. It includes 29 MCP tools, local token usage tracking, save-point backups, and frozen-file editing protection. Scans generate structured Markdown reports formatted specifically for AI coding context.
It suits developers and security professionals using AI assistants who want fast, local static analysis without continuous cloud API fees. It does not suit teams looking for centralized cloud dashboards, non-technical users, or enterprise environments that prohibit humor-based terminal notifications.
Runs native local checks for exposed secrets, process leaks, network bindings, git history leaks, and container escapes.
Evaluates technical debt, test coverage, dead code, complexity, and doc hygiene using Tree-sitter AST parsing.
Exposes endpoints for code search, save points, file freezing, and diagnostics over stdio and HTTP/SSE.
Tracks Claude Code API spending, cache efficiency, and budget forecasts by reading JSONL diaries locally.
Provides a retro PyQt5 desktop interface with 8 pages, popup alerts, audio feedback, and wizard modes.
Free and open source under the MIT license. Scanning runs entirely locally with zero token cost. The only potential expense is optional LLM API charges if you enable external tip providers using your own API keys.
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
These reviewed AI tools do the same job instantly — from our scored directory.
No. All core scanning modules run locally on your machine. API calls are only made if you explicitly configure an optional AI provider key for custom advice tips.
It exposes an MCP server over stdio or HTTP/SSE, which connects directly to any MCP-compliant editor host.
It supports Anthropic, OpenAI, DeepSeek, OpenRouter, NVIDIA NIM, HuggingFace Inference, and Azure OpenAI via config.toml or environment variables.
See Zekai first in Google