Codeplane is an experimental fork of opencode, an open-source AI coding agent designed to run on your own hardware. It offers a self-hosted solution with zero telemetry, ensuring your data remains private. Codeplane supports a wide range of AI models, including OpenAI, Anthropic, OpenRouter, Ollama, vLLM, and custom endpoints, allowing you to choose the best model for your needs. It features a multi-surface architecture, providing front-ends for terminal, desktop (Electron), web, and iOS. Sessions are synchronized across all your devices on the network, allowing you to seamlessly continue your work from anywhere. The agent supports Model Context Protocol (MCP) servers for enhanced functionality, bundling filesystem and GitHub integrations. Codeplane also offers advanced features like queued follow-ups for task management, session branching and sharing via deep-links, a plugin SDK for custom tools and agents, and permissioned approval rules for enhanced security.
Self-hosted: Runs as a server on your machine, VPS, or homelab with no Codeplane cloud dependency.
Any Model Support: Integrates with OpenAI, Anthropic, OpenRouter, Ollama, vLLM, and custom OpenAI-compatible endpoints.
MCP-native: Wire in any Model Context Protocol server, including filesystem and GitHub integrations.
Multi-surface: Access your coding agent from terminal, desktop, web, and iOS devices.
Queued Follow-ups: Stack tasks, reorder them, and manage sessions with branching and deep-link sharing.
Plugin SDK: Extend functionality with custom tools, agents, and prompts using @codeplane-ai/plugin.
Permissioned: Features per-directory and per-session approval rules, with an option for global auto-accept.
Codeplane is built with privacy as a core principle, storing all session data locally in a SQLite file. It offers a flexible and powerful coding experience tailored for developers who prioritize control and privacy.