Whetlan

The .tsr export is the feature I didn't know I needed. I've been manually copying MCP configs between my work laptop and a home server, and it's exactly the kind of thing that breaks silently when you forget one entry. Quick question on multi-client sync: if I have Cursor on one machine and Claude Desktop on another, does the sync work across machines, or is it local to a single device for now?

16M downloads on mobile is wild. Curious how the creation flow translates to web. Do people use it differently when they have a full screen and keyboard vs phone?

Curious about one thing: how does GLM-5-Turbo handle tool calls that depend on results from previous calls in the same chain? That's usually where agent models fall apart for me. The model "forgets" what it got back two steps ago and starts guessing. The fact that it's optimized from training instead of bolted on as a fine-tune afterthought is a different bet. Wondering if that actually helps...

Been juggling Claude Code and Codex on the same codebase and they keep stepping on each other's files. So the isolation part — does each agent get its own worktree, or is it more branch-per-task? Curious how the review works when two agents touch the same file.

Interesting that each agent runs in its own isolated Docker sandbox. Makes sense for security, but curious what happens when you need agents to actually talk to each other? Like if I deploy a research agent and a coding agent and want them to coordinate on the same task — is there any built-in way for them to discover and communicate, or would I need to wire that up myself through external APIs?

After dealing with those npm postinstall attacks lately, seeing an agent that actually validates findings is a massive relief. Most tools just spam false positives until you kill the notifications. Does this catch transitive dependency stuff too, or just first-party code?