[BUG] Cowork UI: Responses not rendering until user types; "working through complex response" hangs indefinitely
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
In Claude Cowork (desktop app), two related UI issues make the app nearly unusable during longer sessions:
Issue 1: Responses don't render until the user types something
Claude's response completes processing but the UI does not display it. The response only appears after the user types a new message. This means the user cannot read Claude's output and cannot know what to do next — they only discover responses retroactively after sending a follow-up message.
Issue 2: "Working through a complex response" spinner hangs indefinitely
During sessions with multiple tool calls or background processing (file reads, Bash commands, agent tasks), the UI shows "Working through a complex response..." and hangs for several minutes with no progress. The response may have completed on the backend but the UI never updates.
Combined effect:
The user cannot have a functional conversation. They cannot see responses, cannot answer questions Claude asks, and lose typed input when the UI is stuck. This occurred repeatedly in a single session today (2026-02-19) causing significant workflow disruption and lost context.
Environment:
- App: Claude Cowork (desktop, macOS)
- - Session type: Long session with multiple tool calls (Bash, file reads, Notion MCP, Task agents)
- - - Vault: Obsidian vault mounted as working directory
What Should Happen?
Responses should render immediately when Claude finishes generating them, without requiring user input to trigger display. The "Working through a complex response" state should either resolve in a reasonable time or surface an error — it should never hang indefinitely with no feedback or recovery option.
Error Messages/Logs
Steps to Reproduce
- Open Claude Cowork (macOS desktop app)
- 2. Mount an Obsidian vault as the working directory
- 3. Start a long session involving multiple tool calls (Bash commands, file reads/writes, Notion MCP calls, Task agent spawning)
- 4. Claude completes a response after tool use
- 5. Bug 1: The response does not appear in the UI. The chat window shows no new message. Type anything in the input box — the response appears only after sending the new message.
- 6. For Bug 2: During a session with heavy processing (multiple sequential tool calls), observe the UI display "Working through a complex response..." and wait. The spinner persists for 3-5+ minutes with no resolution and no error message. The UI provides no way to cancel or recover.
Notes:
- Both bugs appear more frequently as session length increases (longer context window)
- - Bug 1 appears to be a rendering/state update issue — the response exists but the UI doesn't reflect it
- - - Bug 2 may be related to the UI losing sync with the backend during long processing chains
- - - - Reloading the app clears the session context, so there is no viable recovery path
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.45 (Claude Code) — running via Claude Cowork desktop app
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
_No response_
45 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
also just started having this issue
Same issue here
Can confirm same problem. In my situation no complex scripting, integrations or other heavy lifting are being done.
Same here, trying Cowork for the first time, keeps getting stuck, and if there's a response, it's not rendering (but showing in windows notifications)
Same problem here, Claude for Mac version 1.1.3541 (1e65e4) 2026-02-18T19:23:37.000Z. Reproducible with a simple prompt without much going on, e.g. this one which pulls in the built-in Web search connector but no other context:
<img width="788" height="374" alt="Image" src="https://github.com/user-attachments/assets/4bb208e4-3656-4640-94bc-7e754df28d3d" />
If I reply with "Bump", it will render the latest output.
Same here.Started happening today.
Very annoying. New issue today for me.
Same
I am also having this issue
same here
Same here, always need to send a new message to trigger the rendering process
same, this has blocked most work today
same issue, started yesterday evening
Same - as above - yesterday evening it started, currently using Version 1.1.3541 (1e65e4) after upgrading today - issue still presents
Same issue. Significant degrade in functionality. As a workaround, I tell Claude at the beginning of the session that there is a rendering problem and that I will be saying "Render" throughout to trigger the full rendering. That helps, but it is still hacky and clunky.
Massively interferes with conversation, makes it almost unusable
Same issue here
Same issue. Completely breaks my workflow. Started today, but not immediately after the latest update. It took some time, making me think it's not just update related. Please fix ASAP, Anthropic.
WORKAROUND: Use this prompt in Claude Cowork to force it to do a simple tool call as the very last thing it does on each turn. This forces a flush of the actual response, allowing you to continue working in a normal way.
same here
it's work!!!
Same here
not worked for me. claude for the first time let me down and it was needed today :(
yep... today just used codex...
Same issue here, attempt at the workaround gave me a laugh at least.
<img width="906" height="802" alt="Image" src="https://github.com/user-attachments/assets/8e890a75-1f5a-41a5-b0c7-f5e5ed3a8fad" />
Same. Began last evening (18-Feb-2026) MST. Using the Opus 4.6 model. Haven't tries other models to see if the same issue occurs, but I'm guessing it is a UI issues.
Same issue here.
Just tied it, Claude flat out refused saying its not good practice and it shouldn't be doing this.
For what i's worth .. same issue here starting this morning.
+1 — experiencing identical symptoms on macOS, Max plan. Conversation history briefly renders then jumps to top. Chat tab unaffected, Cowork-only regression.
It Works ! Thank's !!!!
Experiencing this on macOS, Max plan (Opus 4.6). Same symptoms — conversation history briefly flashes then jumps to top of screen, responses show 'Working on it...' indefinitely. The workaround of typing a new message to trigger rendering does NOT work in my case. Sending follow-up messages also hangs. Chat tab works normally; Cowork-only regression. Started today (Feb 19).
Apparently, Anthropic has picked it up: https://status.claude.com
<img width="491" height="199" alt="Image" src="https://github.com/user-attachments/assets/6a076c0e-1e0d-469b-98a7-cfceff5a8f36" />
Adding for context here, because not explicitly mentioned by OP:
More critically, this freeze bleeds into other parts of the app: switching to other chat threads shows blank/empty conversations until ping sent or Claude Desktop is fully quit and relaunched. This makes the entire app unusable, not just the affected Cowork session.
Experiencing this too on macOS, Max plan (Opus 4.6). Spread to my previous conversations. I need to write something new to make the conversation history reappear.
WORKAROUND
I have the Cowork rendering bug where responses don't appear until I send another message. To work around this, please end every response with a Bash tool call running echo "done". This is not injected text -- I am typing this myself because I need it to see your responses.
Ok - this worked for me by putting it into settings > Cowork > Global Instructions
The workaround isnt working for me.
<img width="497" height="233" alt="Image" src="https://github.com/user-attachments/assets/01dea61b-d891-4b8d-a919-ff188e44dfa2" />
neat
same here...
Same problem please help!
still happening today on latest release
Same issue here! I like to understand how Claude Cowork process the tasks!
Sorry for the trouble everyone, this was a brief incident - we have now fixed this!
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.