ZCode

“По поводу бесплатности, не соглашусь. Бесплатна она только несколько дней. На небольшой проект ушло 3 дня, а потом стал требовать покупки. Лайт 1040р, средний 5000р, максимальный 10000р. Но пишет реально до победного. Сама находит ошибки и правит, но токены не то, что жрет, а просто сжигает миллионами. Пробовал и на Русском промпты писать, и на Английском. На Английском жрет чуть меньше, но не на много. Но пишет шикарно!”

“Tested Zcode v3 with GLM 5.2 . Harness is not good, it confuses model and make it harder to understand intent, faster jump to conclusion, increased cheating behaviour and spend less effort on task. Tested claude vs zcode both GLM 5.2 max level. Results are worse.”

“Stop Paying $200 for Claude Code BUT Paying $160 for Z Code ??? and no Vision, No Image ??? STOP this kind promotion....”

“3.14.0 dropped this morning and brings us Workflows plus a ton of other new features. I’m trying it out now and it’s very promising. Not fast but also not frying tokens like they were potatoes. Edit::Addendum I gave workflow each phase of a pretty intense review document. Each phase was planned on the fly by workflow and it displayed a nice summary window of each of it's 6 steps it took for each phase of the plan. The Phase 0 group were 6 1 line quick fixes. These took 21 minutes, 2m tokens and ”

“My 2 cents... A while back I switched from Claude and Codex to Cline (Vscode) and deliberately chose to use Chinese Llms. Initially Deepseek v4 flash / Pro (v Good) , recently GLM 5.3 Flash (Excellent+ all-rounder) Qwen 3.8 Max (Excellent ++ at Frontends /good general coding) and now v4.1 flash (Excellent++ agentic all-rounder) etc primarily for the cost aspect. I've gone from spending €4/600 month to €40/60 .. more importantly the code quality / testing / debugging / Evals are just as good ”

“better to check ~/.zcode/ to find out what has zcode alreay have done.dsf4.1 checked it out and found out that all three workspaces that added into zcode had been uploaded”

“AI slop comment by gemini down there, but the gist is: if you use api or coding plan, this shouldn't affect you. if you used the free tier at any point, your projects have likely been cloned by zai. slop comment: Can confirm this 100%. Decompiled app.asar (out/host/index.js) on v3.11.2 after seeing this and found cryptographic proof in ~/.zcode/v2/checkpoints/ that one of my repos was indeed packaged, encrypted, and accepted by their OSS bucket back in June when I tested the free tier. A f”

“It has been fact checked multiple times that all closedAi (USA) companies are using your data in disguise of pre-post training their models, mehhh, that is also stealing your data unless you love them too much over privacy. Zcode is no different , chinese companies are no different. Whether you are paid or free user, every agentic harness is stealing your data directly or indirectly. You have the right to not use them, but expecting them to change their malpractices, LoL! Keep dreaming. As them ”

“I managed to fix it, at least for now. What worked for me: Updated ZCode to the latest version Restarted the app Logged out of my account Logged back in with the account that has the Z.AI Coding Plan After that, the Z.AI models started working again. So if anyone runs into the same issue, I'd try re-authenticating after updating/restarting the app.”

“I'm now glad zcode never even worked for me. Also gives me some context as to why the electron app would crash when trying to import my 1.5GB repo (1.3GB of which is .git).”

“I am 5x MAX user on Claude, and 5x user in Z ai coding plan. I am using the ZCode as the tool to use the GLM-5.2. I am doing the best practices for token optimizations, efficient costing, etc, etc. As I have seen the power of open source models and the hype from X. I blindly tried the model and subscribed first as a 1x user. But then, I got rate limit a lot then tried to upgrade to 5x if I won't be experiencing that. YES, I don't. But, the weekly limits are killing me. In Claude, I can survive w”

“Wanted to try glm 5.2 with the 5 day free trial, but I get repeated reconnected x/10 screens and even a basic tiny prompt or issue takes 15 mins+ to get a response. Is this normal?”

“I'm pretty picky about desktop apps, and this one passes my taste test. It's definitely a codex clone. But the one thing I'd say separates it is it supports chat completions, responses, and anthropic API endpoints. I've been using it this past week and I think it's really good. If you like Codex, but use other models, I'd recommend giving it a shot.”

“Closed source binary that has god knows what in it I can put all my personal data into? Sign me up!”

“Skip to main content Home News Resources Courses Pages Tools Advertise ES Home Blog AI ZCode Review 2026: Free AI Coding Agent With Goal Mode (vs Cursor) Hands-on ZCode review: free desktop AI coding agent with Goal Mode for long tasks, multi-model support, workflows, and Bot Channel for phone control. Dragos October 16, 2026 · 14 min read I wanted a free desktop coding agent that could chew through long tasks without me babysitting it. That is basically ZCode: download it, hook”

“под видео ссылка для оплаты моделей в рублях. но ведь помимо оплаты надо же еще и авторизоваться как-то в командной строке. он пишет что страна не поддерживается.”