gpt-5-codex model endpoint support + Codex CLI documentation

[dylbarne]

1. enable compatibility with new gpt-5-codex model, by default copilot-api routes to chat/completions endpoint however looks like codex model expects different endpoint
   • semi related to other issue gpt-5-codex Model Incompatibility in VS Code 1.98.1 #116
2. provide some documentation around compatibility with usage to Codex CLI config (similar to existing docs showing how can be used with claude code settings already)
   • here is minimal working toml config I have been using: (although if repo maintainers have other suggestions around that would be appreciated, if when added to README potentially)

model = "gpt-5-mini"
review_model = "gpt-5-mini"
model_provider = "copilot"

[model_providers.copilot]
name = "Github Copilot"
base_url = "http://127.0.0.1:4141"

[vtomnet]

#118

[caozhiyuan]

@dylbarne @vtomnet in codex cli in done, https://github.com/caozhiyuan/copilot-api/tree/feature/gpt-5-codex . but in claude code ,need more test . use gpt-5-codex generated these codes

[Louis454545]

@dylbarne @vtomnet in codex cli in done, https://github.com/caozhiyuan/copilot-api/tree/feature/gpt-5-codex . but in claude code ,need more test . use gpt-5-codex generated these codes

wow thanks for that, but how to setup copilot-api with codex ?

[caozhiyuan]

@Louis454545 config.toml
model = "gpt-5-codex"
model_provider = "github-copilot"
model_reasoning_effort = "high"

[model_providers.github-copilot]
name = "github-copilot"
base_url = "http://localhost:4141"
env_key = "GITHUB_COPILOT_API_KEY"
wire_api = "responses"

[zicjin]

@caozhiyuan

▌ hello?

⚠️ stream error: unexpected status 404 Not Found: 404 Not Found; retrying 1/5 in 187ms…

⚠️ stream error: unexpected status 404 Not Found: 404 Not Found; retrying 2/5 in 414ms…

⚠️ stream error: unexpected status 404 Not Found: 404 Not Found; retrying 3/5 in 863ms…

I used the same configuration, aren't you getting the same result?

copilot-api console:

<-- POST /responses
--> POST /responses 404 1ms
<-- POST /responses
--> POST /responses 404 0ms
<-- POST /responses
--> POST /responses 404 0ms

[caozhiyuan]

@zicjin #119 I have added the ping event, which currently resolves the codex disconnection issue: Transport error: error decoding response body. I've tested it and it's working fine now.

please update code , use https://github.com/caozhiyuan/copilot-api/tree/feature/gpt-5-codex

[caozhiyuan]

this is my codex config
model = "gpt-5-codex"
model_provider = "github-copilot"
model_reasoning_effort = "high"

[model_providers.github-copilot]
name = "github-copilot"
base_url = "http://localhost:4141"
env_key = "GITHUB_COPILOT_API_KEY"
wire_api = "responses"
request_max_retries = 3
stream_max_retries = 0
stream_idle_timeout_ms = 300000

[guanghechen]

hi @caozhiyuan, I tried your fork branch and it works great. Thank you very much for your work. During usage, I encountered a small issue I'd like to ask you about: after enabling the codex web_search feature, when I run codex --search, I get the following error. Do you have any idea what might be causing this?

unexpected status 400 Bad Request: {"error":{"message":"{\"error\":{\"message\":\"The use of the web search tool is not supported.\",\"code\":\"unsupported_value\",\"param\":\"tools\",\"type\":\"invalid_request_error\"}}\n","type":"error"}}

[caozhiyuan]

@guanghechen copilot don't support web_search tool, you should deny WebSearch. if use in claude code , add this setting https://github.com/ericc-ch/copilot-api/blob/master/README.md?plain=1#L319 . codex also can disable web_search tool

[guanghechen]

Okay, but if we disable the web-search tool, how can the agent fetch the latest documents or information from the web? Doesn't this cause the agent to always use the data it was trained on?

[caozhiyuan]

@guanghechen use mcp fetch or websearch

[guanghechen]

Thank you for your fast reply! Do you have any recommended MCP fetch tools, Context7? By the way, what does "websearch" mean? Is it an MCP tool?

[caozhiyuan]

@guanghechen "context7": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@upstash/context7-mcp"
]
},
"fetch": {
"command": "uvx",
"args": [
"mcp-server-fetch",
"--ignore-robots-txt"
]
}

[dylbarne]

@zicjin #119 I have added the ping event, which currently resolves the codex disconnection issue: Transport error: error decoding response body. I've tested it and it's working fine now.

please update code , use https://github.com/caozhiyuan/copilot-api/tree/feature/gpt-5-codex

@caozhiyuan I have been using your forked /all branch for Codex CLI lately and has been working great for the most part (for Planning/review etc.) although when it comes time to make file edits am running into 500 status internal server error intermittently (with gpt-5.2 model)

Failed to create responses Response { status: 500,
  statusText: 'Internal Server Error',
  headers:
   Headers { 'content-security-policy': "default-src 'none'; sandbox",
     'content-type': 'text/plain; charset=utf-8',
     'copilot-edits-session': '***',
     'strict-transport-security': 'max-age=***',
     'x-content-type-options': 'nosniff',
     'x-request-id': '***',
     date: '***',
     'content-length': '***',
     'x-github-backend': 'Kubernetes',
     'x-github-request-id': '***' },
  body: ReadableStream { locked: false, state: 'readable', supportsBYOB: true },
  bodyUsed: false,
  ok: false,
  redirected: false,
  type: 'basic',
  url: 'https://api.business.githubcopilot.com/responses' }

my config mostly matches the one you gave with respect to provider setup

[caozhiyuan]

Since when has it been like this? I tried locally and it's the same. I've been using claude code. @dylbarne This is a copilot server-side error, possibly related to apply_patch custom_tool_call, it's not a function call.