ShellGate

Turn locally authenticated CLI tools into an OpenAI-compatible REST API.

Log in to a CLI tool once. ShellGate proxies any HTTP client through it — no separate API keys, no extra billing accounts.

Your App  →  POST /v1/chat/completions  →  ShellGate  →  codex exec "..."
(N8N, LangChain, any OpenAI SDK)                      →  kimi --print "..."
                                                      →  claude -p "..."
                                              ↑
                                        your CLI logins

Requests route automatically based on model name — use Codex, Kimi, and Claude simultaneously from the same endpoint.

Supported Providers

Provider	CLI	Model prefix	Auth	Status
OpenAI Codex	codex	gpt-*	codex login	✅ Supported
Kimi CLI	kimi	kimi-*	kimi login	✅ Supported
Claude Code	claude	claude-*	claude login (OAuth)	✅ Supported
Antigravity CLI	agy	—	—	🔜 Planned

Install

One-line (Linux/macOS):

curl -fsSL https://raw.githubusercontent.com/DutaKey/ShellGate/main/install.sh | sh

From source:

git clone https://github.com/DutaKey/ShellGate
cd ShellGate && make build

Quick Start

Option A — Guided setup wizard

shellgate setup

Walks through config creation, provider login, API key generation, and server start in one flow.

Option B — Manual

# 1. Create config (~/.shellgate/config.toml)
shellgate init

# 2. Log in to CLI providers
shellgate login codex
shellgate login kimi
shellgate login claude

# 3. Start the server
shellgate serve -d          # background
# or
shellgate serve             # foreground

# 4. Create an API key
shellgate keys

Use http://localhost:8080/v1 as your OpenAI base URL.

Model Routing

ShellGate routes each request to the right provider based on model name:

Model	Provider
gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.3-codex, gpt-5.2	Codex CLI
kimi-code/kimi-for-coding	Kimi CLI
claude-opus-4-7, claude-sonnet-4-6, claude-haiku-4-5	Claude Code CLI

All providers are always active — no config switch needed.

CLI Reference

shellgate setup                 Guided first-time setup wizard
shellgate init                  Create config file interactively
shellgate login <provider>      Authenticate a CLI provider (codex, kimi, claude)
shellgate serve                 Start the API server (foreground)
shellgate serve -d              Start in background
shellgate stop                  Stop background server
shellgate restart               Restart background server
shellgate status                Show server status + provider auth status
shellgate logs                  View last 50 log lines
shellgate logs -f               Follow log output (tail -f style)
shellgate keys                  Interactive key manager (arrow keys)
shellgate keys create <name>    Create API key (scriptable)
shellgate keys list             List all keys
shellgate keys revoke <id>      Revoke a key

Flags:

-c, --config string   config file (default ~/.shellgate/config.toml)

API

Drop-in replacement for the OpenAI API.

Method	Path	Description
POST	/v1/chat/completions	Chat completions (streaming + non-streaming)
POST	/v1/responses	Responses API (N8N AI Agent, LangChain)
GET	/v1/models	List available models (all providers)
GET	/v1/models/:id	Get model by ID
GET	/health	Health check
POST	/admin/keys	Create API key
GET	/admin/keys	List API keys
DELETE	/admin/keys/:id	Revoke API key

Usage Examples

curl — Codex:

curl http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer <your-key>" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.4","messages":[{"role":"user","content":"hello"}]}'

curl — Kimi:

curl http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer <your-key>" \
  -H "Content-Type: application/json" \
  -d '{"model":"kimi-code/kimi-for-coding","messages":[{"role":"user","content":"hello"}]}'

curl — Claude:

curl http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer <your-key>" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-sonnet-4-6","messages":[{"role":"user","content":"hello"}]}'

Python (OpenAI SDK):

from openai import OpenAI

client = OpenAI(
    api_key="<your-shellgate-key>",
    base_url="http://localhost:8080/v1"
)

# route to Claude
response = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "hello"}]
)

# route to Codex
response = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "hello"}]
)

N8N: OpenAI node or AI Agent node → set Base URL to http://<host>:8080/v1.

LangChain / other frameworks: Any client that supports custom OpenAI base URL works.

Configuration

Config lives at ~/.shellgate/config.toml by default.

[server]
host = "0.0.0.0"
port = 8080
read_timeout = "30s"
write_timeout = "120s"

[auth]
admin_secret = ""                          # required — protects /admin/* endpoints
keys_file = "~/.shellgate/keys.json"

[executor]
codex_binary = "codex"
default_sandbox = "read-only"              # read-only | workspace-write | danger-full-access
timeout = "120s"
kimi_binary = "kimi"
claude_binary = "claude"
working_dir = ""                           # override working dir for all CLI tools

[logging]
level = "info"    # debug | info | warn | error
format = "json"   # json | text

Environment variable overrides:

Variable	Config field
SHELLGATE_PORT	server.port
SHELLGATE_ADMIN_SECRET	auth.admin_secret
SHELLGATE_KEYS_FILE	auth.keys_file
SHELLGATE_EXECUTOR_CODEX_BINARY	executor.codex_binary
SHELLGATE_EXECUTOR_KIMI_BINARY	executor.kimi_binary
SHELLGATE_EXECUTOR_CLAUDE_BINARY	executor.claude_binary
SHELLGATE_LOG_LEVEL	logging.level

Docker

docker run -d \
  -p 8080:8080 \
  -v $HOME/.shellgate:/root/.shellgate \
  -v $HOME/.codex:/root/.codex:ro \
  -v $HOME/.kimi:/root/.kimi:ro \
  -v $HOME/.claude:/root/.claude:ro \
  -e SHELLGATE_ADMIN_SECRET=your-secret \
  ghcr.io/dutakey/shellgate:latest

License

MIT