Just 3 minutes, connect RuToken to Codex in one click

I want to connect the  RuToken API to Codex. I am not technical, so please figure everything out yourself and do not ask me technical questions. Just explain the result in plain language when you are done.

── My details ────────────────────────────
 RuToken API key: <paste your key here>
Preferred model: <leave blank if unsure, pick one for me>
──────────────────────────────────────

Please follow these 6 steps exactly:

[STEP 1] Check the environment
Run codex --version and note the version number.
Locate the config file:
  macOS / Linux → ~/.codex/config.toml
  Windows       → %USERPROFILE%\\.codex\\config.toml

[STEP 2] Back up first
If config.toml already exists, make a backup copy with today's date in the filename (e.g. config.toml.bak-20261012).
If the folder or file does not exist, create it.

[STEP 3] Verify the key BEFORE touching any config
Use curl to call https://rutoken.ai/v1/models with the header: Authorization: Bearer <my key>

  Works        → show me the list of available model names (names only, do not paste raw JSON)
  401 or 403   → stop and tell me "the key is invalid, please copy it again from the  RuToken dashboard". Do not modify the config.
  Network down → tell me the exact error. Do not modify the config.

Never print the key to the output and never write it into any log or temporary file.

[STEP 4] Pick a model and actually test it
Choose a model suitable for coding from the list in step 3, preferring the gpt-5 family.
Then test it with a minimal curl request to https://rutoken.ai/v1/responses:
  body: {"model":"<model name>","input":"hi"}
If it errors (model not found, endpoint unsupported, etc.), try the next one in the list, up to 3 attempts.
Tell me which model finally worked.

[STEP 5] Write the config
Write parts A and B below into config.toml. Three hard requirements:

  (a) Part A contains TOP-LEVEL keys. They must go at the very beginning of the file, before any [bracketed] section.
      If they end up after a [section] header they are silently ignored and Codex quietly falls back to
      the official endpoint — this is the single most common cause of failure.
  (b) Part B goes at the very end of the file.
  (c) Keep all existing content intact (e.g. [tui], [mcp_servers]). If the file already contains model or
      model_provider lines, only change their values — never write them twice. Duplicate keys make Codex
      fail to start.

  ── A: at the very top of the file ──
model = "<the model name from step 4>"
model_provider = "newapi"

  ── B: at the very end of the file ──
[model_providers.newapi]
name = " RuToken"
base_url = "https://rutoken.ai/v1"
wire_api = "responses"
experimental_bearer_token = "<my key>"

Three things to be careful about:
  · base_url must be exactly https://rutoken.ai/v1 — do not append /responses (Codex adds it)
  · wire_api must be "responses". Setting "chat" makes Codex fail to start
  · Do not rename or add any of these fields. A misspelled field is ignored without an error

[STEP 6] Self-check and report back
Run codex doctor and confirm the Configuration and auth sections are healthy, especially:
  config.toml parse   ok
  wire API            responses

Then tell me four things:
  1) The full path of config.toml on my machine
  2) Which model is now configured
  3) This reminder: "Completely quit Codex and reopen it for the new config to take effect"
  4) If doctor reports any fail or warn, explain in plain language what the problem is and what I should do

Please do not show me code or JSON at any point.