Engineered for software engineers who demand speed, complete model flexibility, and explicit permission control. Works with Anthropic, AWS Bedrock, OpenAI, DeepSeek, OpenRouter & Ollama.
Single binary installation with no node or python runtime required. Run the one-line installer for your operating system.
Open Terminal and execute the script below. It detects Apple Silicon (M1/M2/M3/M4) or Intel automatically:
curl -fsSL https://heksync.in/install.sh | bash• Installs binary atomically into ~/.local/bin/heksync
• Adds ~/.local/bin to your shell configuration (.zshrc / .bash_profile)
We built Heksync Code to eliminate vendor lock-in, unconstrained background actions, and token budget bloat.
Never worry about an agent executing destructive shell commands or clobbering files unannounced. Choose your safety boundary per turn:
Switch seamlessly between cloud models and local offline LLMs without changing tools or configuration formats.
$ heksync --provider bedrock
$ heksync --provider anthropic
$ heksync --provider ollama
Long-running debugging sessions can consume hundreds of thousands of tokens. Heksync Code monitors history density and automatically compacts preceding turns into structured summaries before limits are hit.
Your code, prompt history, session transcripts, and SQLite database stay entirely inside your machine at ~/.heksync. No telemetry or hidden tracking.
Upgrading your CLI is as simple as running one command:
Fetches the latest binary, checks the detached SHA-256 digest, and atomically swaps the executable.
When models stream truncated JSON or hit token output limits mid-turn, our parser automatically repairs unclosed quotes and brackets without crashing your active session.
Connect to any frontier cloud LLM or run completely offline with local models.
Direct API support with prompt caching enabled for massive token savings.
Uses Bedrock Converse API with SigV4 signing across any AWS region.
OpenAI-compatible endpoints, Azure OpenAI, DeepSeek, and custom proxy gateways.
Access over 100+ models with customizable reasoning effort and fallbacks.
100% offline development. No internet connection or API keys required.
Layered config support: Global (~/.heksync/config.toml), Project (.heksync/config.toml), and ENV.
Powerful terminal workflow with interactive commands and non-interactive scripting options.
/initGenerates an AGENTS.md file with project context & rules./modeSwitch permission mode (ask, acceptEdits, plan, yolo)./modelSwitch LLM model or provider on the fly./planPropose a structured step-by-step implementation plan./todoView and manage the session task list./clearClear current screen buffer while keeping session saved.Install Heksync Code now and experience fast, secure, autonomous AI coding built for developers.
Install Heksync Code