Product

Secure your AI coding agents without changing how developers work

Route Claude Code, Codex, and other agent CLIs through Outgate Gateway. Protect secrets, enforce guardrails, choose providers, scan projects, and give teams controlled agentic workflows from the terminal.

Connect existing developer tools to your Outgate AI Gateway with one command. Use your own providers, routers, regions, and guardrail policies while keeping credentials and sensitive data out of upstream model calls.

$og login
$og region change eu-central-1
# Route Claude Code through Outgate
$og claude
# Route Codex through Outgate
$og codex --provider "OpenAI"
# Scan a repo before agents touch it
$og scan --provider "Smart Router"

Repositories, prompts, logs, and tool output can contain secrets, PII, credentials, and sensitive internal context. Each agent also brings its own provider setup, authentication behavior, and routing quirks.

og CLI lets teams keep the developer tools they already use while routing agent traffic through Outgate's protected AI gateway for guardrails, provider control, scanning, and regional governance.

og CLI gives teams a control layer for terminal-based AI agents. Developers can run og claude or og codex to route Claude Code and Codex through Outgate while keeping their normal workflow.

Teams can use configuration to define repo-specific providers, routers, regions, auth mode, scan settings, and private gateway URLs. They can also run og scan to detect PII, credentials, secrets, and sensitive project data before real model requests are made.

For production environments, og CLI supports Anthropic, OpenAI, local models, smart routers, failover routing, SSH-friendly paste login, CI tokens, private deployments, and multi-region gateway routing.

Built for the teams that ship AI

Engineering teams

Let developers use Claude Code and Codex while enforcing shared routing, guardrails, and provider policies.

Platform teams

Standardize AI agent access across repos, regions, CI jobs, local machines, and private deployments.

Security teams

Detect sensitive data, manage vault entries, reduce credential exposure, and keep model traffic behind controlled gateway policies.

Regulated environments

Support private regions, custom endpoints, self-hosted gateways, and auditable AI workflows.

How it works

From terminal agent to protected model request

  1. 01Developer runs og claude, og codex, or og scan
  2. 02og CLI resolves config from flags, .og.yaml, env vars, and global settings
  3. 03The CLI selects the right region, provider, router, share, and gateway URL
  4. 04Prompts, files, and agent traffic pass through Outgate Gateway
  5. 05Guardrails anonymize or block sensitive data before upstream model calls
  6. 06Requests route to Claude, OpenAI, local models, or your own backend LLM
  7. 07Responses return to the original CLI workflow

Drop og CLI into your developer workflow

Install in seconds, point it at any region, and start protecting Claude Code and Codex traffic today.

Frequently asked questions

It wraps existing AI agent CLIs and routes their traffic through Outgate Gateway for provider routing, guardrails, scanning, and protected execution.
No. Developers keep using those tools. og CLI configures the environment and gateway path around them.
Yes. Use og login --token-file with a pre-generated developer token, then run og claude, og codex, or og scan in non-interactive workflows.
Yes. Add .og.yaml to the repo to pin providers, regions, shares, gateway URLs, auth mode, and scan behavior.
No. The scan uses the guardrail service in dry-run mode. Detections are stored in the Detection Vault and the request never reaches the upstream provider.