Introducing HyCode

We built HyCode because we were tired of choosing between two extremes: fully manual coding or handing everything off to an AI that doesn't understand your workflow. HyCode is a terminal-native UI for hybrid coding — you and your AI agents work together, in the same terminal, on the same codebase, at the same time.

The idea is simple. You stay in control of the architecture and the hard decisions. Your agents handle the grunt work — scaffolding, refactoring, test generation, boilerplate. You review, approve, and steer. It's pair programming where your partner never gets tired and scales to as many parallel tasks as you need.

HyCode runs entirely in your terminal. No browser tabs, no Electron apps, no context switching. The TUI is built for speed: syntax-highlighted diffs, inline approvals, real-time streaming output, and a layout that keeps everything visible without clutter. If you spend your day in the terminal, HyCode feels like home.

One of the things we're most proud of is model agnosticism. HyCode supports 13+ LLM providers out of the box — OpenAI, Anthropic, Google, Mistral, Cohere, Ollama, Together, Groq, and more. Swap models mid-session. Use a fast local model for quick edits and a frontier model for complex refactors. Your API keys, your choice.

The sub-agent system is where things get interesting. Spawn parallel agents that work on different files or different parts of a problem simultaneously. One agent writes tests while another implements the feature. A third agent reviews the diff. They coordinate through a shared context layer so nothing conflicts. For large refactors across dozens of files, this is a game-changer.

We also ship with full MCP (Model Context Protocol) support. MCP lets agents share structured context — file trees, dependency graphs, type information — without stuffing everything into a prompt. It means better results with less token waste, and it means your agents actually understand the shape of your codebase rather than guessing from a few files.

The skills system rounds things out. Skills are reusable workflows — think of them as recipes for common tasks. "Add a REST endpoint with tests and types" can be a single skill invocation. You can build your own, share them across projects, or pull from the community marketplace. Skills compose with agents, so a single command can orchestrate a complex multi-step workflow.

HyCode is free for local models. The Pro plan unlocks all cloud providers, unlimited parallel agents, the full skills marketplace, and priority support for $19/month. Teams get shared skills libraries, SSO, and custom model endpoints.

We've been using HyCode internally for months, and it's changed how we write software. The feedback loop is tighter. The tedious parts disappear. And because everything happens in the terminal, there's no friction — just code. We're excited to share it with you.