Changelog
New updates and improvements to OpenCode
v1.18.2 Improved default reasoning depth for Meta models
Fixes
- CoreStopped subagents from launching nested subagents by default, with a configurable
subagent_depthlimit when needed.
- CoreStopped subagents from launching nested subagents by default, with a configurable
v1.17.20 Updated Azure AI support for GPT-5.6
Fixes
- CoreRemoved an obsolete Codex workaround that could interfere with OpenAI Luna Responses Lite requests.
v1.17.19 Supported OpenAI pro reasoning mode
Fixes
- CoreDisabled response storage by default for xAI Responses.@geraint0923
- CoreAdded OAuth support for Luna Responses Lite.
- CoreSwitched to another available org after logging out in the console.
- CoreUsed Codex context limits for GPT-5.6 over OAuth.@nabilfreeman
- TerminalForwarded CLI environment variables to the TUI worker.
v1.17.18 Add a model-specific system prompt for Meta Muse Spark
Fixes
- CorePrevent crashes and bad pricing data when GitHub Copilot returns models with a zero billing batch size.
v1.17.17 v1.17.16 Exposed reasoning effort variants for Grok models
Fixes
- CoreImproved xAI prompt cache routing and PDF file support in Responses models.
v1.17.15 Handle unavailable config directories more gracefully when reading config files
Fixes
- CoreBetter classify Z.ai context-window overflow errors so oversized requests surface the right failure mode@fengjikui
v1.17.14 Added a code mode MCP adapter for running confined orchestration scripts against connected MCP tools
Improvements
- CoreHid the
executetool unless code mode is enabled.
Fixes
- CoreFixed paginated MCP tool catalogs losing tool metadata and output schema validation.
- CorePreserved low reasoning effort for OpenRouter small-model variants instead of disabling it.
- CoreFixed GitHub Copilot model routing to honor each model's advertised chat or responses endpoint.
- CoreFixed session lists to match equivalent instance directories reliably.
- CoreFixed Cerebras reasoning replay so earlier assistant reasoning is sent back in the provider-supported field.
- TerminalFixed spinner registration so loading indicators keep rendering across TUI surfaces.
- CoreHid the
v1.17.13 Stop replaying stale GitHub Copilot response item IDs, avoiding follow-up request failures
Fixes
- CoreForce reasoning mode for OpenAI-compatible reasoning models so reasoning settings apply reliably on custom deployments
v1.17.12 Add a yolo mode to auto-approve permissions
Improvements
- TerminalPass ServerAuth headers through external served TUI connections.@OpeOginni
- SDKAdd a live event subscription stream.
- SDKAdd SDK access to active sessions.
- SDKRestore session runtime operations such as event streaming, interrupts, and message lookup.
- SDKAdd paged durable session history.
- SDKAdd session permission request create and fetch endpoints.
Fixes 13
- CoreEnable adaptive thinking for Claude Sonnet 5.
- CorePrefer MCP content responses over structured output when both are present.
- CoreReconnect MCP servers after OAuth even if the server was disabled.@MaxAnderson95
- CoreRequest MCP refresh-token scope during OAuth.
- CoreShow MCP OAuth completion errors instead of a generic failure.
- CoreRefresh cached remote skills.
- CorePreserve skill resource paths.
- CorePick better default small models across providers.
- CoreScope MCP auth status to each server URL.
- SDKPreserve
V2Eventnames in SSE streams. - SDKWake embedded session execution after new prompts.
- SDKImprove V2 runtime tool results and avoid duplicate model switch events.
- SDKResolve MIME types for prompt attachments sent by URI.