Announcing YottaCode — An Autonomous AI Coding Agent for Your Terminal

Today we’re excited to announce the public release of YottaCode — an autonomous, self-learning AI coding agent that runs directly in your terminal.

What is YottaCode?

YottaCode is a command-line AI agent that helps you code, debug, and ship software from your terminal. Unlike typical coding assistants that just autocomplete or chat, YottaCode takes action — it reads your codebase, writes files, runs tests, manages git, opens PRs, and learns from every interaction.

Key capabilities

• Autonomous multi-step tasks — Give it a goal (“add OAuth login”, “refactor the auth module”, “fix the flaky test”) and it plans, implements, tests, and iterates
• Self-learning memory — Plain-text, grep-able memory that gets sharper the more you use it. You own USER.md, the agent curates the rest
• Use any model, switch any time — Native adapters for Anthropic, OpenAI, Gemini, xAI, Ollama, plus ChatGPT/Copilot OAuth. Swap mid-session with /model <name>
• Built-in security — Per-workspace trust prompts, approval modals with syntax-highlighted diffs, allow/ask/deny rules, and write-path validation that firewalls secrets
• Deep GitHub integration — Typed GitHub adapter (no gh CLI needed). Read issues, open PRs, review code, and take issues end-to-end
• Plan mode — Read-only investigation that drafts a plan you approve before any code changes
• Rewind & resume — Every prompt auto-checkpoints. Double-tap Esc to rollback. Sessions save atomically; /recall searches all history
• Parallel worktrees — Run multiple agents on the same repo without collisions
• Extensible — Custom slash commands via markdown files, MCP server support, reusable skills

Getting started

# Install on Linux or macOS
curl -fsSL https://yottacode.ai/cli/install.sh | bash

# Run the interactive setup wizard
yottacode setup

Watch the YottaCode setup walkthrough:

Built for developers, by developers

YottaCode is free and open source (MIT license). It’s a single static Go binary — no runtime dependencies, no telemetry, no hidden defaults that silently bill you.

We built it because we wanted an AI agent that:

• Works in the terminal where we already live
• Doesn’t lock us into one model provider
• Respects our security boundaries
• Learns our codebase and preferences over time
• Can actually do the work, not just suggest it

Join us

• GitHub: yottadynamics/yottacode — star, watch, contribute
• Documentation: yottacode.ai/docs
• Issues: Report bugs or request features
• Community: Follow @YottaDynamics and @WayOfMandalore_ for updates

Thanks to everyone who tested early versions and gave feedback. Let’s build something great together.

— The YottaDynamics Team