Is there a way to sandbox the agent ?
Is there a way to restrict terminal command from a agent to access/edit file outside the current directory ? I see that gemini-cli or codex-cli use seatbelt on macOS. I don't see any equivalent for copencode.
OpenCode known issues
Known breaks in OpenCode from github.com/anomalyco/opencode, with quoted workarounds and fixed-in versions where they exist.
Before you upgrade OpenCode
Type the version you're about to install (or just upgraded to). Lists tracked issues that mention it. Same data as /api/issues/opencode.json.
Mentioned in open issues:
Pulled from public GitHub issues updated in the last 45 days and filtered for always-on problems: install and update breakage, crashes and hangs, gateway/daemon trouble, auth, and headless setups. Ranked by engagement, weighted toward always-on, self-hosting, install/update, crash, auth, MCP and daemon problems (cosmetic desktop-UI issues are down-weighted). Workarounds and "fixed in" versions are quoted only when the issue or its top comments state them. Severity is an automatic estimate from labels, keywords, and engagement, not an official rating. Last refresh: Oct 26, 2026 6:18 AM ET.
OpenCode v1.17.10 crashes with Bun segmentation fault on Windows; v1.17.9 appears stable
After upgrading OpenCode to v1.17.10, OpenCode started crashing with a native Bun segmentation fault on Windows. Downgrading to v1.17.9 appears to stop the fatal crash under the same repository/configuration/session pattern. This looks…
Workaround quoted from issue body
Downgrading to `opencode-ai@1.17.9` appears to avoid the fatal crash under the same environment and workflow.
always stuck at “Preparing write...”
My Prometheus always stuck at “Preparing write...”。
desktop: custom provider save always throws "unavailable on this server"
OpenCode Desktop exposes the "Custom OpenAI-compatible provider" form in Settings → Providers, but the save handler unconditionally throws provider.custom.unavailable, so the flow can never succeed on any server — including the app's own…
Forbidden, OpenCode Free Model Gateway Detect Abuse Traffic
Forbidden, OpenCode Free Model Gateway Detect Abuse Traffic
Regression in 1.18.30: every prompt crashes with TypeError in SystemPrompt.environment ("a.name") — 1.18.18 works fine
After updating to opencode 1.18.30 via pacman, every single prompt fails immediately with: Failed to send prompt. Unexpected server error. Check server logs for details. This happens on a brand new session, on the very first message, with…
Workaround quoted from issue body
Downgrading to 1.18.18 (via the cached pacman package) fixes it immediately — same machine, same everything else.
Unable to install OpenCode on windows with winget command
Unable to install OpenCode on windows with winget command I follow the instruction described in Changelog: https://github.com/anomalyco/opencode/blob/dev/CHANGELOG.md#highlights-3 The root cause…
migrate-v1: non-git sessions in "global" project hidden after V1→V2 upgrade
After upgrading from V1.18.31 to V2.0.8, sessions that V1 stored in the global project (sessions created in non-git directories) become invisible in the TUI session picker, because V2 creates a per-directory project and the picker filters…
Workaround quoted from issue body
Before repair: 104 sessions total, 93 with `project_id = 'global'` (81 in `/home/wjf`, 8 in `/tmp/opencode`, 3 in `/home/wjf/novel`, 1 in `/home/wjf/videos`). - Sessions in a git repository were unaffected: `/home/wjf/fadebreak` kept a stable project id across the upgrade and stayed visible. -…
SystemPrompt.environment crashes on every prompt: undefined is not an object (evaluating 'a.name')
OpenCode fails before sending any prompt to the model. The UI shows: The server log shows: The error occurs repeatedly for new sessions and different working directories. Observed output: macOS 27.0(Apple Silicon)
OAuth Authorization Fails with HTTP 400 in OpenCode Desktop 2.0.28 on Windows
OpenCode Desktop 2.0.28 fails to complete ChatGPT OAuth authorization on Windows. After selecting the ChatGPT sign-in option and proceeding through the browser authorization flow, the browser displays Authorization failed with the message…
[gateway] 30-image cap applied only to some session routes: identical request flips 400/200 with only x-opencode-session changed; capped sessions cannot /compact
The OpenCode Go gateway (https://opencode.ai/zen/go/v1) enforces a 30-image-per-request cap on only some routes: enforcement follows the x-opencode-session routing value and the model. Identical requests (same API key, same model, same…
Workaround quoted from issue body
Workaround used: temporarily switch the model/provider, compact there, switch back. - The same request works or fails purely based on the session header, which is extremely hard for users to diagnose. - Compounds #47493 (no client-side media strip) and #39677 (image-count 400 not classified as…
openai: ChatGPT OAuth requests use Zen API key
incorrectly binds the OpenAI provider to the OpenCode Zen integration, causing a Zen API key to be sent to ChatGPT’s OAuth-only Codex endpoint. OpenAI rejects the key with “API keys are not supported by this endpoint.” The correct OpenAI…
Free usage limit does not reset after 24+ hours - stuck in infinite retry loop
I am using the free tier of OpenCode. I received the "Free usage exceeded, subscribe to Go [retrying in XXXXXs]" message. The problem is that the limit does NOT reset even after waiting for more than 24 hours and doing nothing in the app…
NO API KEY
I subscribed to OpenCode Go today. My Go subscription is active, but the console does not show any personal API key or any option to create/copy one. /connect → OpenCode Go requires an API key. The Keys page only allows creating Service…
Incorrect API key provided: sk
OpenCode returns an OpenAI 401 when sending a prompt. The error is: `Incorrect API key provided: sk-svcac OpenCode’s bot is asking you to edit the issue body (not just comment). The numbered fields were left as No response. You have about…
Go API key is not available in the new v2 Console
I have an active OpenCode Go subscription, and Go models work normally in OpenCode CLI. However, I cannot find any way to create, view, or copy a Go API key in the new v2 Console. The current Go documentation still says users can copy an…
tui: false "Session location unavailable" prompt after network reconnect
When the TUI reconnects after a brief network interruption, it can replace the composer with the "Session location unavailable — Choose another directory to continue this session" recovery prompt even though the session's directory is…
run: --command removed in v2 and slash commands are not expanded — no CLI path for headless command invocation
OpenCode 2.0.8 removed --command from opencode run, and run does not expand slash commands in the message, so there is no CLI way to invoke a command headlessly. --command was the supported headless path in v1 — see #7345, where the…
terminal does not clipboard read (kitty OSC 5522) for image paste over SSH
I use the TUI over SSH and can't paste images. Unsure if it's a bug or a gap in the design: the clipboard I want to read is on my laptop, but the TUI is running on the remote box. Add a read path to the terminal clipboard backend using…
Models get stuck on "Thinking" after a few hours
After using OpenCode for a few hours, any model I try gets stuck indefinitely on "Thinking". There is no error message and the request never finishes. Switching to another model does not help. I have not reached any usage or rate limits…
providers: GitHub Copilot model sync never runs after login, no models listed
GitHub Copilot OAuth login succeeds and the account is stored, but no Copilot models ever appear in /models or opencode models. The model-list sync is never attempted after login: the server log contains no sync success or failure for the…
opencode-go/deepseek-v4.1-flash randomly rejects requests with 400/403
I'm hitting random failures with deepseek-v4.1-flash on the opencode-go provider. The session just stops with: Bad Request: {"model":"deepseek-v4.1-flash"} Sometimes I also see it come back as a 403 with error code: 1010 when I replay the…
VSCode plugin broken
On start it use opencode --port 16955 Unrecognized flag: --port in command opencode but correct syntax is opencode serve --port 16955 opencode v2.0.10 Install opencode & vscode plugin Use button with black square on white "Open opencode in…
`opencode auth login` crashes on Linux without xdg-open
On a Linux box without xdg-utils (SSH, containers), opencode auth login prints the authorization URL and then exits with Error: Executable not found in $PATH: "xdg-open" and a stack trace, so the login can't be finished. openUrl in…
[Bug] Mistral provider (zai-glm-5-2) fails with AI_TypeValidationError on streamed tool calls (@ai-sdk/mistral dependency update needed)
OpenCode Version: 1.18.29 Model / Provider: Mistral (mistral/zai-glm-5-2) When using Mistral models (specifically zai-glm-5-2) with tool calling enabled, streaming responses crash with an AITypeValidationError. The error trace indicates…
tui: /btw crashes when rendering answer dialog
Stop letting @thdxr push :) Running the built-in /btw slash command crashes the TUI after the generated answer is returned. The answer dialog attempts to access PluginProvider outside its context boundary. opencode version: 2.0.8 Shell…
Android/Bionic: Bun 1.4.1 splitting:true causes a.name crash
Totally not AI-generated OpenCode 1.18.31 on Android/Bionic crashes before the provider call when compiled with Bun 1.4.1 and splitting:true. Observed failure: TypeError: undefined is not an object (evaluating 'a.name') Affected endpoints…
Workaround quoted from issue body
`minify:true, splitting:false`
No issues match those filters.