Uh oh!
There was an error while loading. Please reload this page.
feat: let ACP client expose the input/output properly - #11303
feat: let ACP client expose the input/output properly#11303validatedev wants to merge 19 commits into
Conversation
The following comment was made by an LLM, it may be inaccurate: No duplicate PRs found |
f1ae801 to
08fa7f7Comparerekram1-node
commented
Jan 30, 2026
Can you explain the changes, it looks like quite a bit |
soar
commented
May 15, 2026
What are we waiting for? This needs to be fixed |
validatedev
commented
May 15, 2026
@rekram1-node@nexxeln@thdxr waiting to be reviewed, just fyi. |
soar
commented
May 21, 2026
How is it going here? |
kian1991
commented
May 27, 2026
Yeah pushing this. I am checking in every day, i guess there are just too many open PRs. |
Steen3S
commented
May 28, 2026
spotlesscoder
commented
May 28, 2026
please prioritize this |
…pers - Remove parse-command.ts in favor of logic consolidated in tool-format.ts - Rewrite tool-format helpers (toolCallFromPart, toolResultFromPart, permissionDisplayInfo, fenceWith) as the single source of truth for tool call formatting - Adapt agent.ts to dev's post-namespace-refactor layout (flat exports with `export * as ACP` self-reexport, ProviderID/ModelID branded types, Hash.fast, InstallationVersion, ConfigMCP.Info) - Drop listSessions `unstable_` prefix to match SDK method name - Guard sendUsageUpdate against missing providerID/modelID
makoMakoGo
commented
Jun 6, 2026
when to merge? the current opencode acp is totally not useable. I cant bear it anymore. why the opencode maintainers even didn't leave any reviews. I want to say that opencode is totally dog shit. 😡 |
kian1991
commented
Jun 10, 2026
Bumping here as well please review |
…run-command-message # Conflicts: # packages/opencode/src/acp/event.ts # packages/opencode/src/acp/tool.ts
…message' into fix/acp-show-proper-run-command-message
soar
commented
Jun 12, 2026
I assume this will never be fixed |
KraXen72
commented
Jun 13, 2026
Hello, any progress on this? Just found the original issue and it seems to have been auto-closed due to inactivity. OpenCode maintainers, can you please review this and merge it? thanks! |
validatedev
commented
Jun 14, 2026
Done and merged to dev thanks to @nexxeln :) |

What does this PR do?
Fixes#10998.
Old behavior: we sent a
tool_callatpendingwithkind: "execute". As the tool ran, updates carried the content plus a completedrawOutput. Zed relied onkind: "execute"to render the blue "run command" box, but the actual command (rawInput) only appeared after completion, so you couldn't see what was running.New behavior: the first
tool_callnow reflects the tool part's actual status (e.g.in_progresswhen the tool is already running on first observation) and carries the fullrawInput— including the resolved working directory forbash— so Zed can show the command and where it's running before output lands. Subsequenttool_call_updates fill in streaming content and the finalrawOutput(stdout/stderr).kindis set toother(notexecute), which removes the blue run-box styling and instead exposesrawInput/rawOutputas collapsible panels.How did you verify your code works?
By running Zed's ACP inferface via:
Screenshot
Old Behavior
New Behavior