local-agent - AI-Powered Code Analysis Tool

🔍

Smart File Scanning

Recursively scans your project with intelligent filtering. Supports source files, configs, PDFs, PCAP captures, and more. Respects .gitignore and custom deny/allow patterns.

💬

Interactive Chat Mode

Full terminal UI and web UI to chat with your codebase. Live rescan capability, focus mode to target a single file, and session logging built in.

⚡

Concurrent Batch Processing

Process large codebases fast with configurable parallel LLM requests. Tune AGENT_CONCURRENT_FILES to match your Ollama setup.

🔒

Privacy First

100% local processing via Ollama — no data ever leaves your machine. Also supports remote Ollama instances via --host for team setups.

📊

PCAP & PDF Analysis

Parse and analyze network traffic captures (.pcap, .pcapng) and extract text from PDF files up to 10MB — all within the same familiar interface.

📦

Standalone Binary

Single binary with all assets embedded. No runtime dependencies beyond Ollama. Ships for macOS and Linux (arm64 + amd64).

✨ Why local-agent?

🤖 Any Ollama Model Works with gemma, deepseek, mistral and any model you have pulled locally
🎯 Task-Oriented Prompts Pass a -task flag for one-shot analysis or drop into interactive mode for multi-turn conversation
🌐 Remote Ollama Support Point to a remote ollama instance — great for shared GPU servers
🔧 Highly Configurable Token limits, concurrency, model, temperature, filters — all tunable via config file
📋 Session Logging All chat sessions are automatically logged for later review and auditing
🛠️ Built for Developers Focus mode, model switching at runtime, health check, model listing — everything you need at the CLI