OpenCode v2.3.0

OpenCode v2.3.0 was released on .

Summary

OpenCode 2.3.0 is here: subagents now run quietly in the background while you keep working, transcripts collapse into tidy activity rows, and you can jump to any earlier prompt in the desktop app. Curious what else changed?

Release Context

  • This was the 211th OpenCode release of 2026 tracked by Havoptic.
  • It shipped 3 days after OpenCode v2.2.7 (November 10, 2026).
  • OpenCode: ships every 2.1 days — 13 releases in the last month.

Why it matters: Subagents now work in the background while you keep coding.

Shipped around the same time

Within 3 days of this release, other AI coding tools also shipped:

Compare OpenCode

Release Notes

Highlights - Background Subagents by Default: Subagents now return immediately, continue in the background, and report back whenever they become idle. You can keep working in the parent session while multiple subagents run concurrently. (63382) - Quieter Transcripts by Default: The TUI now summarizes each run of tool calls and thoughts into one collapsible activity row, so the answer is easier to follow. To see every step again, set Verbosity to medium or high in settings, or set session.verbosity in your TUI config; an existing setting is kept. (63207) - Jump to an Earlier Prompt: In the desktop app, open Jump to message from the command palette or press Cmd+Alt+G (Ctrl+Alt+G on Windows and Linux) to search prompts across the current session and jump directly to one. (63585) Core Improvements - Subagents now always run in the background, return their session ID immediately, and report results to their parent whenever they become idle. (63382) - Added experimental support for MCP events, behind the mcpevents flag. (63254, 63261, 63269) - Shell commands now survive server handoffs and keep their output when interrupted. - Added experimental support for moving a shell command to the background when its timeout expires, behind the shelltimeoutbackground flag. With the flag on, an explicit timeout on a background command stops it. (63024, 63703) - Local Windows shell commands now run through the process daemon. - Added support for freeform tool input on compatible models. - Added LM Studio decision-model support. - Agents are now told when their session is a fork. Thanks @hasta84. (63235) - Tool output truncation now keeps the beginning of the output and preserves a usable preview if the full output cannot be saved. - Tail reads no longer load the entire file into memory. - Providers can present a client certificate for gateways that require mutual TLS. Set settings.tls with cert, key and an optional passphrase on the provider; values accept {file:} and {env:} like apiKey. (61832) - When a subagent goes idle while its own background work is still running, its report to the parent now says it is still running, so the parent waits for the final report instead of treating the partial answer as done. (63675) - Code Mode gets simpler structured results from file and skill tools: write and edit return their confirmation as a string, patch returns an empty result, and skill returns the skill's directory and content. (63616, 63635, 63618, 63623) Bug fixes - Fixed case-only file renames in patches on case-insensitive filesystems. - Fixed stale thinking being retained after Claude Agent prompts or tools change. - Fixed directly driven child sessions not resuming after a restart. - Fixed plugin replacement stopping when cleanup of an earlier plugin fails. - Fixed MCP OAuth refreshes and localhost redirects when a server rewrites 127.0.0.1. - Fixed saved API keys being omitted from local-model discovery. - Local model discovery (Ollama, LM Studio, vLLM) waits up to 15 seconds per request instead of 1, so remote or tunnelled endpoints list their models, and a slow endpoint no longer delays other endpoints or config reloads. (63700) - In Code Mode, calling a namespace as if it were a tool now suggests up to three real tools under it. (63697) - Fixed retryable provider streams losing completed work when interrupted. - Fixed GitHub Copilot requests using the wrong endpoint before models finish loading. - Fixed exhausted request rate limits being treated as terminal. Thanks @Kyle-j-bush. (59816) - Fixed LM Studio models with a null context length and accepted decision-type models. Thanks @tomzhu1024. (56214) TUI Improvements - Transcript verbosity now defaults to low, which summarizes each run of tools and thoughts into one activity row. A session.verbosity you already set is kept. (63207) - Session IDs found in messages can now be opened directly. - Activity tabs are clickable and open the correct view. - Shell mode has its own prompt color under the new designs. Bug fixes - Restored pasted text when undo brings back its placeholder. - Restored the live draft after browsing prompt history. - Fixed copying a paste from within the prompt selection. - Preserved the Mini composer caret when closing the command palette. - Prevented image bytes from being retained in transcript storage when image previews are disabled. - Fixed paste-chip contrast across light and dark themes. - Fixed plugin-provided themes updating live. - Session lists now open at the top rather than at the current session. - Mini scrollback now shows the subagent card instead of obsolete launch instructions. Desktop Improvements - Added Jump to message to the command palette with Cmd+Alt+G / Ctrl+Alt+G, including prompts not...

View original release

Other Recent OpenCode Releases

All OpenCode releases · All AI coding tool releases on Havoptic