opencode
Your terminal's AI agent, with any model you want
5.0•50 reviews•1.5K followers
Your terminal's AI agent, with any model you want
5.0•50 reviews•1.5K followers
opencode is an open-source AI coding agent built for the terminal. It features a native TUI, multi-session support, and is model-agnostic, letting you use Claude, OpenAI, Google, or even local models.
This is the 3rd launch from opencode. View more
OpenCode 2.0
Launched this week
OpenCode is an open source AI coding agent. It’s available as a terminal-based interface, desktop app, or web app.
Free
Launch Team
Dropbox SignSimple, secure eSignatures for the way your team works.
Promoted
Codex by OpenAI
Claude Code
Google Antigravity
That's so cool. Great work, OpenCode team. :)
Niceee! Good luck today
Really like working with OpenCode. Thanks for a great tool.
Congrats! Can i run 6 sessions with one account with a single provider credential (for example an Anthropic API key, or OpenCode Zen), can all 6 sessions run concurrently? Are there per-account limits on your side, or is it purely the provider's rate limits?
Congrats on 2.0! I'm firmly in the "the harness matters more than the model" camp, so model-agnostic is the right bet. Curious how sessions handle the context wall now. Do you checkpoint and resume, or still summarise once you get near the limit? Lossy summaries right before the finish line were the #1 way my agents broke.
Congrats to Fang and the OpenCode team on 2.0. Shipping TUI, desktop, and IDE as first-class surfaces is a real step up from terminal-only agents. For teams that bounce between laptop and remote box, do MCP servers, permissions, and session history sync across those surfaces, or is each install still its own silo today?