sdlc.cc

Pre-release. sdlc.cc is not generally available. There is no completed customer deployment and no design partner yet, and per-class detection accuracy has not been published. What is and is not proven →

JOURNEY_SENTINEL_CODEX · Preview

Codex gateway
Privacy before OpenAI sees the prompt.

A native installer sets Codex openai_base_url to https://api.sdlc.cc/v1 — not api.openai.com. The gateway scrubs OpenAI Chat Completions and Responses bodies, then forwards upstream. Preview: scrub-forward, not full durable enforcement yet.

Download installers → Guided setup

Why Codex needs its own path

Codex talks OpenAI’s API (/v1/responses), not Anthropic Messages. Browser Guard never sees those calls. Pointing at OpenAI directly bypasses sdlc.cc.

Not a base-URL myth anymore

Older docs said OPENAI_BASE_URL at the gateway could not work. Exact /v1/responses and /v1/chat/completions now scrub-forward when the host has OPENAI_API_KEY.

Installer writes Codex config

Sets ~/.codex/config.toml openai_base_url and your sk_sdlc_* key in the user environment — not a wiki of env vars.

Still Preview

Unlike Claude Code’s Supported /v1/messages lifecycle, this path is scrub-forward with a Preview banner header. Durable enforcement parity remains Planned.

How it works

Codex CLI / IDE │ │ openai_base_url=https://api.sdlc.cc/v1 │ OPENAI_API_KEY=sk_sdlc_* (messages scope) │ ▼ sdlc.cc · POST /v1/responses or /v1/chat/completions │ scrub prompt fields · refuse images/files ▼ OpenAI upstream (host OPENAI_API_KEY) │ ✕ never direct to api.openai.com from the laptop

Install

PlatformArtifactWhat you do
Windows SdlcCodexGatewaySetup-windows-amd64.exe Double-click → enter key → quit/reopen Codex → -verify
macOS …-macos-arm64.app.zip Open the .app → enter key → quit/reopen Codex
Linux linux-amd64 binary Run in a terminal 
# Silent
SdlcCodexGatewaySetup-windows-amd64.exe -silent `
  -api-key sk_sdlc_... -gateway https://api.sdlc.cc/v1

# Verify
… -verify
# openai_base_url must be https://api.sdlc.cc/v1 — never api.openai.com

Legacy /v1/completions and query-string variants of OpenAI routes still return UNSUPPORTED_MODEL_WRITE. Images/files are refused. Signed MSI / notarized DMG remain Planned.

Coverage

SurfaceStatusBoundary
Codex via installer → sdlc /v1/responses Preview Scrub-forward; host needs upstream OPENAI_API_KEY; stream:true refused
OpenAI Chat Completions via gateway Preview Same scrub-forward contract; non-streaming only
Cloudflare AI Gateway worker Preview Alternate path; still valid
ChatGPT web Browser Guard Supported Different product — not Codex
Durable enforcement parity with /v1/messages Planned Required before claiming Supported
Uploads, images, voice, legacy completions Unsupported Fail closed

Get Codex on the gateway

  1. Download the installer from /downloads/.
  2. Activate a sk_sdlc_* key with the messages scope.
  3. Run the installer, quit Codex fully, reopen, then -verify.
  4. Confirm the gateway host has OPENAI_API_KEY for upstream.
Download Codex gateway → Open guided setup

Preview software. Claude Code’s Supported path is documented separately on the integrations page.