The Wire · Showcase
Cline desktop plugins crash on launch
By RepoJournal · Filed · About AI Agents
14 people shipped this
Saoud Rizwan
@saoudrizwan
2 cited
langchain-oss-model-profiles
@langchain-oss-model-profiles
1 cited
JatinJain703
@JatinJain703
1 cited
mkondratek
@mkondratek
1 cited
UjjwalRana12
@UjjwalRana12
1 cited
KhanCold
@KhanCold
1 cited
langchain-oss-model-profiles[bot]
@langchain-oss-model-profiles[bot]
kodurd
@kodurd
Badar Rahman
@DarRahman
smolpaws
@smolpaws
enyst
@enyst
VascoSch92
@VascoSch92
johnwschoi
@johnwschoi
LancyZhao
@LancyZhao
Cline's desktop users hit a wall installing marketplace plugins, and the fix is a rewrite of the installer that skips the missing CLI.
Cline shipped a fix for desktop marketplace installs [1] that were failing because packaged apps inherit launchd's minimal PATH on macOS and most users have no cline CLI. The new code installs plugins and MCP servers in-process, eliminating the 'Executable not found in $PATH: "cline"' error. Separately, Cline cut a severe memory leak in hub events [2] where snapshots captured the entire message transcript, ballooning a session to 25GB on a 16GB Mac. The fix stops shipping full transcripts inside broadcast hub events. Also fixed: apply_patch no longer rewrites CRLF files to LF [3], which was blowing up diffs on Windows. Over in llama_index, StructuredLLMRerank gained native async support [4] so concurrent batches actually run in parallel, and the MCP streamable-http client now matches mcp 2.0's 2-tuple API [5]. OpenHands validated API keys before advancing the backend connection step [6], and LangChain refreshed model profiles [7], adding Qwen3.8-27B with 262K context.
One email a day. Unsubscribe in one click.
Keep up with AI Agents in about 3 minutes a day: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
One email a day. Unsubscribe in one click. Read a past issue →
Action items
- → Update Cline desktop to the latest build to get the in-process installer and transcript fix cline/cline [immediate]
- → Upgrade llama_index to include async StructuredLLMRerank and the MCP unpack fix run-llama/llama_index [plan]
- → Monitor Cline for CRLF fixes before your next Windows diff review cline/cline [monitor]
References
- [1] fix(desktop): install marketplace plugins and MCP servers in-process instead of spawning a cline binary ↗ cline/cline
- [2] fix(hub): stop shipping full transcripts inside broadcast hub events ↗ cline/cline
- [3] fix(tools): preserve a file's own CRLF line endings across apply_patch updates ↗ cline/cline
- [4] feat(core): add native async support to StructuredLLMRerank ↗ run-llama/llama_index
- [5] fix(tools/mcp): correct streamable-http tuple unpack for mcp 2.0 (#22655) ↗ run-llama/llama_index
- [6] fix: validate api key before advancing backend connection step ↗ OpenHands/OpenHands
- [7] chore(model-profiles): refresh model profile data ↗ langchain-ai/langchain
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?v=4&s=48 "github-actions[bot]"){kind=small}
