codex新手使用问题
- 内容介绍
- 文章标签
- 相关推荐
转用GPT模型了,在claude code上使用了几天感觉工作流与opus区别挺大,而且压缩上下文有时候会卡住不动,所以想换到codex。
API渠道,win环境,使用codex推荐哪种客户端?
感觉和cc不同,桌面端和vscode插件比cli用人的更多?我还是希望能用cli
codex的插件和claude code通用吗?
希望尽量无痛移植,沿用claude code的配置
康康佬们codex配置
网友解答:--【壹】--:
关于客户端选择,我个人在windows端的使用习惯是WSL2 + codex cli + windows terminal(我实际上在win环境也是做linux开发)
--【贰】--:
mcp
-groksearch
-acetools
-context7
windows推荐cli+pwsh7
mac看个人喜好
skill看需求,我自己搓了一套针对前端设计/语言风格/输出结构的整合包
--【叁】--:
codex就用codex cli
Pycharm用Codex Luncher也是cli
插件应该不太通用吧,至少superpowers的配置方式就不一样
配置文件献丑了
model_provider = "api"
model = "gpt-5.4"
model_reasoning_effort = "xhigh"
disable_response_storage = true
approvals_reviewer = "user"
plan_mode_reasoning_effort = "xhigh"
model_instructions_file = "./instruction.md"
[model_providers.api]
name = "api"
base_url = "http://xxxx/v1"
wire_api = "responses"
requires_openai_auth = true
[mcp_servers]
[mcp_servers.chrome-devtools]
type = "stdio"
command = "cmd"
args = ["/c", "npx", "-y", "chrome-devtools-mcp@latest"]
startup_timeout_ms = 20000
[mcp_servers.chrome-devtools.env]
PROGRAMFILES = 'C:\Program Files'
SystemRoot = 'C:\Windows'
[windows]
sandbox = "unelevated"
[notice]
hide_full_access_warning = true
[features]
multi_agent = true
guardian_approval = true
prevent_idle_sleep = true
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\brainstorming\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\writing-plans\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\executing-plans\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\dispatching-parallel-agents\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\requesting-code-review\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\receiving-code-review\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\using-git-worktrees\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\using-superpowers\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\finishing-a-development-branch\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\systematic-debugging\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\test-driven-development\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\verification-before-completion\SKILL.md'
enabled = false
--【肆】--:
推荐pwsh,alt +/- 分屏用起来也很爽
IMG_04781920×1043 211 KB
仅演示分屏
--【伍】--:
总感觉codex cli回复格式和布局排版看着很难受,所以用插件更多,感觉跟cli也没什么区别
转用GPT模型了,在claude code上使用了几天感觉工作流与opus区别挺大,而且压缩上下文有时候会卡住不动,所以想换到codex。
API渠道,win环境,使用codex推荐哪种客户端?
感觉和cc不同,桌面端和vscode插件比cli用人的更多?我还是希望能用cli
codex的插件和claude code通用吗?
希望尽量无痛移植,沿用claude code的配置
康康佬们codex配置
网友解答:--【壹】--:
关于客户端选择,我个人在windows端的使用习惯是WSL2 + codex cli + windows terminal(我实际上在win环境也是做linux开发)
--【贰】--:
mcp
-groksearch
-acetools
-context7
windows推荐cli+pwsh7
mac看个人喜好
skill看需求,我自己搓了一套针对前端设计/语言风格/输出结构的整合包
--【叁】--:
codex就用codex cli
Pycharm用Codex Luncher也是cli
插件应该不太通用吧,至少superpowers的配置方式就不一样
配置文件献丑了
model_provider = "api"
model = "gpt-5.4"
model_reasoning_effort = "xhigh"
disable_response_storage = true
approvals_reviewer = "user"
plan_mode_reasoning_effort = "xhigh"
model_instructions_file = "./instruction.md"
[model_providers.api]
name = "api"
base_url = "http://xxxx/v1"
wire_api = "responses"
requires_openai_auth = true
[mcp_servers]
[mcp_servers.chrome-devtools]
type = "stdio"
command = "cmd"
args = ["/c", "npx", "-y", "chrome-devtools-mcp@latest"]
startup_timeout_ms = 20000
[mcp_servers.chrome-devtools.env]
PROGRAMFILES = 'C:\Program Files'
SystemRoot = 'C:\Windows'
[windows]
sandbox = "unelevated"
[notice]
hide_full_access_warning = true
[features]
multi_agent = true
guardian_approval = true
prevent_idle_sleep = true
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\brainstorming\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\writing-plans\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\executing-plans\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\dispatching-parallel-agents\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\requesting-code-review\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\receiving-code-review\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\using-git-worktrees\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\using-superpowers\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\finishing-a-development-branch\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\systematic-debugging\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\test-driven-development\SKILL.md'
enabled = false
[[skills.config]]
path = 'C:\Users\jingt\.codex\superpowers\skills\verification-before-completion\SKILL.md'
enabled = false
--【肆】--:
推荐pwsh,alt +/- 分屏用起来也很爽
IMG_04781920×1043 211 KB
仅演示分屏
--【伍】--:
总感觉codex cli回复格式和布局排版看着很难受,所以用插件更多,感觉跟cli也没什么区别

