CODEX CONFIG BUILDER

Codex Config Generator: Build config.toml Safely

Generate a Codex CLI config.toml and an auth.json placeholder locally in your browser. The page never accepts an API key, and authentication output always contains YOUR_API_KEY.

Data processing: Processed locally on this page
Content updated: 2026-07-19

CONFIGURATION FLOW

Keep configuration and authentication separate

Codex uses config.toml for models, providers, and transport behavior. auth.json carries authentication fields only, so the generator never mixes a real secret into ordinary configuration.

01
Public inputs
Endpoint, model ID, reasoning effort, and WebSocket capability
02
Behavior config
Produce a reviewable ~/.codex/config.toml
03
Auth placeholder
auth.json always contains YOUR_API_KEY only
04
Local validation
Write the key securely in a terminal, then run strict doctor

The page handles public connection settings only; secrets stay out of inputs, URLs, browser storage, and copied templates.

INPUT

Generate public configuration parameters

There is no API key field. Never paste a real key into the endpoint or model fields; authentication output always uses a fixed placeholder.

Codex provider parametersAPI base URLUse a complete http:// or https:// URL without credentials, query parameters, or a fragment.Provider model IDgpt-5.6-sol is a CodeNodex provider ID and may not exist on another endpoint.Reasoning effortGenerate the WebSocket capability fieldEnable only when the provider explicitly supports Codex Responses over WebSocket.

The generator does not read existing files. Back up before a manual write; CodeNodex CLI is safer for automatic merge and validation.

OUTPUT

Review generated output

A real secret never appears in the output.

model_provider = "OpenAI"
model = "gpt-5.6-sol"
review_model = "gpt-5.6-sol"
model_reasoning_effort = "xhigh"

[model_providers.OpenAI]
name = "OpenAI"
base_url = "https://token.codenodex.com/v1"
wire_api = "responses"
requires_openai_auth = true

[features]
goals = true

The configuration text comes only from current page state.

What the Codex config generator outputs

The generator produces two text blocks: Codex CLI config.toml and an auth.json containing a fixed placeholder. config.toml defines an OpenAI Responses-compatible provider, so the model ID and endpoint must exist in the selected service catalog.

The CodeNodex preset uses the current public endpoint and the gpt-5.6-sol provider model ID. This is a CodeNodex routing identifier and must not be assumed to exist on the official OpenAI endpoint or another provider.

Never collects an API key
Never writes local files automatically
Never overwrites existing Codex settings
Never sends form values to an API

The OpenAI and Anthropic configuration boundary

This Codex CLI template targets an OpenAI Responses-compatible provider. Anthropic Messages is a different request protocol; Claude Code and similar clients require their documented configuration rather than a provider-name substitution.

If a service advertises only OpenAI Chat Completions or Anthropic Messages compatibility, verify the wire_api and tool-call behavior required by the current Codex client before connecting it.

SAFETY BOUNDARY

Complete four checks after copying the template

The output is a minimal draft and does not read or merge existing Codex files. Back up before writing and validate with the local CLI afterward.

01
Back up
Keep recoverable copies of config.toml and auth.json
02
Merge
Do not overwrite existing MCP, profile, or security settings
03
Inject
Replace placeholders only in local auth files or a controlled secret workflow
04
Verify
Check parsing, provider, model, and a read-only request separately

For production use, prefer CodeNodex CLI backup, merge, and validation; use the web tool to inspect structure or perform a manual migration.

Continue with another tool

API Cost CalculatorSeparate input and output charges with an explicit formula, then compare the budget for one task, one session, or a batch of requests.Open tool Compatibility MatrixInspect client, protocol, version baseline, and verification scope so a text response is not mistaken for complete workflow compatibility.Open tool

NEXT STEP

Move the local result into a real development workflow

Read the matching client guide, then create a separate API key and complete a minimal read-only verification.

View integration guides Sign up and create a key

Generate public configuration parameters

There is no API key field. Never paste a real key into the endpoint or model fields; authentication output always uses a fixed placeholder.

The generator does not read existing files. Back up before a manual write; CodeNodex CLI is safer for automatic merge and validation.

model_provider = "OpenAI" model = "gpt-5.6-sol" review_model = "gpt-5.6-sol" model_reasoning_effort = "xhigh" [model_providers.OpenAI] name = "OpenAI" base_url = "https://token.codenodex.com/v1" wire_api = "responses" requires_openai_auth = true [features] goals = true

What the Codex config generator outputs

Never collects an API key

Never writes local files automatically

Never overwrites existing Codex settings

Never sends form values to an API

The OpenAI and Anthropic configuration boundary

If a service advertises only OpenAI Chat Completions or Anthropic Messages compatibility, verify the wire_api and tool-call behavior required by the current Codex client before connecting it.

Keep configuration and authentication separate

Public inputs

Behavior config

Auth placeholder

Local validation

Generate public configuration parameters

Review generated output

What the Codex config generator outputs

The OpenAI and Anthropic configuration boundary

Complete four checks after copying the template

Back up

Merge

Inject

Verify

Continue with another tool

Move the local result into a real development workflow

Keep configuration and authentication separate

Public inputs

Behavior config

Auth placeholder

Local validation

Generate public configuration parameters

Review generated output

What the Codex config generator outputs

The OpenAI and Anthropic configuration boundary

Complete four checks after copying the template

Back up

Merge

Inject

Verify

Continue with another tool

Move the local result into a real development workflow