OpenCode

OpenCode is an open source coding agent that works on your own machine and meets you on the surface you prefer: a desktop app with sessions, tabs, diffs, a built in browser, and extensions on macOS, Windows, and Linux, a fast terminal UI that works in any shell, editor, or remote server, a mobile app for starting, steering, and approving sessions from your phone, and an SDK for driving OpenCode from your own scripts, CI jobs, and tools. The project describes itself as the open source coding agent that works where you do, and its homepage reports, as vendor stated figures, more than 220,000 GitHub stars, more than 1,000 contributors, more than 20,000 commits, and use by more than 22 million developers every month. Getting started is deliberately low friction. You download OpenCode and use the free models it comes with, with no account needed, and then add more models when you need them. OpenCode says it connects to AI models from more than 200 providers, including Claude, GPT, and Gemini, and you can bring subscriptions you already have, logging in with a ChatGPT Plus or Pro subscription or with GitHub Copilot, or use an OpenCode subscription instead. Tooling extends the same way: you connect the services you already use through MCP servers, hook into OpenCode with plugins, and teach agents your workflows with skills, with GitHub, Linear, Figma, Sentry, Notion, Slack, Supabase, and Stripe named on the homepage as examples of connected services. The commercial side runs through the OpenCode Console, which the homepage describes as high quality inference for frontier and open weight models, low cost subscriptions with generous limits, and the tools a team needs to manage spend and collaborate. Its Go pricing page describes the model subscription qualitatively as a low cost monthly plan built on curated, affordable models tested for agentic coding, with generous limits and reliable access, plus a higher Plus tier with everything in Go, higher limits, and fewer interruptions for larger, demanding projects. The listed Go models are open weight and frontier adjacent names such as Kimi, DeepSeek, Claude Haiku, Qwen, and GLM variants, with allowances counted over five hour, weekly, and monthly windows that vary by model. The pricing FAQ explains that free models carry a daily request limit, that Go users who reach a limit can keep using the free models or add credit and turn on extra usage so requests draw on that balance instead of stopping, that an owner or admin can cancel at any time with the plan staying active to the end of the paid period, and that Go also works with other coding agents that send typical coding agent traffic with their own user agent and a session ID. On privacy, the FAQ says training and retention terms vary by model and that Go blocks paid endpoints that train on your requests by default, a setting an owner or admin can change.