OpenCode v2.2.5

OpenCode v2.2.5 adds keyboard shortcuts for picking a worktree and its base branch, lets the compaction buffer be set as a percentage of each model's context window, and stops disconnecting MCP servers that send stdio messages over 10 MiB.

Why it matters

For developers using the open-source OpenCode coding agent across several models. One percentage setting now keeps room free in both small and large context windows, and big MCP responses no longer drop the connection.

What's new

• Set "compaction": { "buffer": "20%" } in opencode.json to keep a share of each model's context window free; token counts still work.
• Desktop shortcuts: Cmd+Alt+O (Ctrl+Alt+O on Windows and Linux) picks a worktree, Cmd+Alt+B (Ctrl+Alt+B) picks its base branch, and Cmd+Shift+X (Ctrl+Shift+X) toggles shell mode.
• MCP servers that send stdio messages larger than 10 MiB stay connected, and large messages are read much faster.
• Clearer errors: a failed read names the missing folder and the nearest one that exists, and a failed edit shows the file's closest lines.
• LiteLLM context-window errors now trigger compaction instead of stopping the session, and reasoning is no longer turned into plain text when you switch models.

Taken from the project's own release notes. Claims about speed or accuracy are the maker's unless stated otherwise.

Worth knowing: Calls to a tool named bash now run as shell with the same permissions, unless a plugin tool is actually named bash.

Quick facts

Links

• Official site: https://opencode.ai
• Repository: https://github.com/anomalyco/opencode
• Release notes: https://github.com/anomalyco/opencode/releases/tag/v2.2.5
• Docs: https://opencode.ai/docs

Related ranking

See how established tools compare in Best AI Coding Tools for Developers. New releases aren't scored until we review them against the rubric.

Checked 2026-10-31 from the release notes's own pages. Prices and star counts change, so confirm on the official site. Listing here is not a ranking or an endorsement.