Hi @everyone! We're working on improving onboarding and the overall experience in Traycer Desktop, and we'd love your input. Reply in this thread with anything that felt confusing, took a while to figure out, or could just work better. Every bit of feedback helps us decide what to fix first. P.S.: A lot of you have been asking for the mobile app TestFlight, so here's the link: https://testflight.apple.com/join/GrMtbuXk

.

For me the agent usage/credit usage meter is the most importance area to improve. I am using Kimi and there is no meter for that, can make it customizable for our AI to code/add to the current meter list?

It seems like Traycer eats a lot of RAM and sometimes the artifact does not sync. Maybe add a button to re-sync it. I saw the A2A cross-host is coming. I would really like this feature. Also add an android Traycer version so I can view and continue prompting the agents on the phone. Sorry, this is such a long list 😅

Hey gang, as you know, I'm obsessed with Traycer, so I gave your question directly to SWE2 with some of my wish list items. My top concerns from my memory were #2 and #6. I knew the AI had more on file: > Heavy Traycer user here — day one with a multi-agent estate task surfaced a lot. Biggest things: > > 1. Agents have no goal/brief field. In a 50-agent task I can't query "why does this agent exist" — it's buried in first messages. Also

agent.roles.claim

errors

E_AGENT_NOT_FOUND

for agents that list fine, so the role registry can't be written. > 2. Model providers are confusing. My ZAI sub is usable but hidden under "OpenCode" — the copy is clunky and there's no direct submenu. And Traycer's devin catalog hides the ~70

fusion-*

pairings the CLI exposes, then rejects them server-side if you paste an id. > 3. Worktrees need an overhaul — listing hangs mid-scan on large estates, deletes fail with no reason code on top-level dirs, and the worktree RPC wedged entirely (twice) under a delete batch. File org in general could use a pass. > 4. Artifact dirs reap live files. Payload files I write mid-task get silently flattened or deleted — it killed a run by eating

run_manifest.json

. Please either let artifacts hold files or tell me when a write won't persist. > 5. Agent messaging is silent-failure-prone.

ACP -32011

drops mid-task, and agents finish work without replying — no way to tell dead from thinking. A `last_seen`/turn-state field would fix orchestration. > 6. Wishlist: better skills out of the box, and stronger built-in steering so agents use more of the system without me hand-holding. > Overall love the direction — these are the sharp edges I hit, roughly in pain order.

Nothing to be sorry about! We love detailed feedback. Keep it coming ❤️

Not so avid user but i use your service for extended spec driven planning i do the code implementation myself. one hurdle i have currently is i am not able to fully migrate my current workflow from the traycer vs extension to the desktop app. also where ever any plans to migrate the extension to the zed editor ? Amazing work nonetheless i am truly happy with your service!