Tingly Box CLI
Tingly Box runs a local model gateway: Claude Code, Codex and OpenCode use one address and requests go to the providers you choose
- 381 ★ on the repository
- Last commit
- MPL-2.0
Broad access
Why this level
- The gateway sees every agent request and stores provider keys and OAuth tokens
- Remote control through Telegram and other messengers sends commands to an agent that has access to your project
- Subscription OAuth sign-in in a third-party client may break the vendor's terms and get the account blocked
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/tingly-boxDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The gateway sees every agent request and stores provider keys and OAuth tokens
- Remote control through Telegram and other messengers sends commands to an agent that has access to your project
- Subscription OAuth sign-in in a third-party client may break the vendor's terms and get the account blocked
This is third-party code. Review the repository files before installing.
What it does
Tingly Box is a local gateway written in Go with a web UI on port 12580. The agent talks to one address, the gateway translates Anthropic and OpenAI protocols in both directions and sends the request to the providers you configured. Claude Code and OpenCode get one-click setup, and Xcode is configured by hand with in-app hints. API keys and OAuth sign-in both work, and smart routing picks a model by cost, speed or your own rules. Beyond the gateway the program has team keys with per-member usage tracking, an MCP gateway, guardrails, an image API on the same address and remote control of agents through Telegram, Weixin, WeCom, Feishu, Lark and DingTalk. For Claude Code you can create profiles and switch models for different tasks. Install with npm install -g tingly-box@latest, start with tb start and open the web UI with tb open. You can also run it through npx tingly-box@latest or the Docker image ghcr.io/tingly-dev/tingly-box.
Who it is for. Teams and solo developers who run Claude Code, Codex and OpenCode through different model providers and want one entry point with usage tracking.
Good fit when
- Claude Code has to switch between models for different tasks through profiles
- A team shares model access and each member needs a separate key with separate statistics
- The agent speaks the Anthropic protocol while the provider offers only an OpenAI-compatible API
- You want to drive an agent from your phone through Telegram or another messenger
Not a fit when
- One model from one provider is enough: the gateway adds an extra process
- You need a hosted service: Tingly Box runs on your machine or in your Docker
- You only need protocol translation inside application code: the gateway is a separate process
Example request
Install Tingly Box, start the gateway and connect Claude Code to it with a separate profile for heavy tasksLimitations
- Xcode and clients without a ready template are configured by hand with in-app hints.
- Installing through npm needs Node.js, and the Go binary for your platform comes down with the package.
- The detailed compatibility table in the README covers DeepSeek: cache, reasoning mode, vision and web search. For other providers their API sets the feature list.
- Quick start links in the README point to issue tracker comments rather than to separate documentation.
How to disable. End the tingly-box process and remove the package with npm uninstall -g tingly-box. Remove the Docker container with docker rm -f tingly-box. In each agent restore the original API address that one-click setup replaced.
FAQ
How do I install Tingly Box?
Run npm install -g tingly-box@latest, then tb start and tb open. npx tingly-box@latest works without installing, and the Docker image ghcr.io/tingly-dev/tingly-box listens on port 12580.
How do I connect Claude Code to Tingly Box?
The web UI at http://localhost:12580 has a one-click Claude Code setup button. Profiles for switching models by scenario live in the same place.
How does Tingly Box differ from Claude Code Router?
Claude Code Router picks the model provider and routes requests, while Tingly Box adds team keys with usage tracking, an MCP gateway, an image API and remote control through messengers.
Which keys and providers does Tingly Box accept?
The gateway works with API keys and with OAuth sign-in. The detailed compatibility table in the README covers DeepSeek: Anthropic and OpenAI protocols, reasoning mode, vision, web search and a 1M token context.
How do I hand out keys to a team?
In the web UI open Team, set the model rules and issue keys under Team Keys. Each member's usage appears under Dashboard, Team usage.
Discussion
No questions or reviews yet.
Related
CLI toolsAgentBox runs Claude Code, Codex and Open Code in separate virtual machines on your PC, a self-hosted server or a cloud.
Runs each agent stage in its own micro-VM: Claude, Codex, Ollama or any CLI, with OTLP tracing
ClaudeBox runs Claude Code in Docker with language profiles, session slots and a firewall with an allowlist of addresses.
setup-gateway writes an OpenAI-compatible gateway into the Claude Code, OpenCode and Codex CLI configs in one pass, with a backup.
The GoModel gateway puts MCP servers behind one protected address and keeps their tokens out of the agent's reach.
ExitBox runs Claude, Codex, Cursor and other agents in rootless containers: a domain allowlist and secrets in an encrypted vault.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Tingly Box (this card) | yes | 381 | |
| Claude Code Router | no | 38,127 | a local gateway for provider choice and routing of agent requests |
| routatic-proxy | no | 981 | a local proxy for Claude Code with routing rules |
README in short
The README lists gateway features: any agent with any model, team management, add-on modules and performance. It then covers installation through npm, npx and Docker, an agent integration guide, team keys, the image API, DeepSeek compatibility, remote control through messengers and an OpenAI SDK example.