Setup

Install CodeSeeX, copy generated TOML, restart Codex.

The generated adapter configuration is machine-specific. Prefer copying it from the desktop manager instead of hand-writing paths or ports.

Install the desktop app

On Windows, use the NSIS Windows-CodeSeeX.Setup.*.exe installer from GitHub Releases for normal install and update.

Start the local proxy

Confirm the manager shows the local service running, usually on 127.0.0.1:8787.

Copy the generated TOML

Use the adapter card so model_catalog_json and base_url match the current machine.

Restart Codex and select a model

Choose deepseek-v4-pro for quality or deepseek-v4-flash for speed.

Generated TOML example

This is a structure example only. Copy the real TOML from the CodeSeeX desktop manager because model_catalog_json, base_url, and local paths are machine-specific.

model_provider = "custom"
model = "deepseek-v4-pro"
disable_response_storage = true
model_reasoning_effort = "xhigh"
model_catalog_json = "<generated path to model-catalog.json>"

[model_providers.custom]
name = "DeepSeek"
wire_api = "responses"
requires_openai_auth = true
base_url = "http://127.0.0.1:8787/v1"

Troubleshooting checklist

Most setup failures are catalog path, port, proxy, or tool exposure issues.

If models do not appear, confirm model_catalog_json points to an existing model-catalog.json.
If requests fail, confirm Codex base_url points at CodeSeeX and the port is not occupied.
If web search is weak, check source diagnostics and system proxy mode.
If tools behave unexpectedly, confirm the tool is enabled and supported by the current Codex session.

GitHub Releases

Use the NSIS installer for desktop install, update, and legacy migration.

Download from Releases Security