AI AgentAI AgentPlugins And Skills
CC Switch
Open source, cross-platform AI programming configuration management tool
标签:AI Agent Plugins And SkillsPlugins and SkillsWhat is CC Switch?
CC Switch is an open-source, cross-platform AI programming configuration management tool that supports macOS, Windows, and Linux. As a one-stop AI programming tool hub, CC Switch provides unified management of vendor configurations, local routing, usage tracking, Skills synchronization, and cross-application session recovery for seven applications: Claude Code , Claude Desktop , Codex , Gemini CLI , OpenCode , OpenClaw , and Hermes Agent , all within a single interface. It enables integration with third-party models with zero code modification.
Main functions of CC Switch
- Unified vendor management : A single interface manages the model vendors, authentication information, and API endpoints for seven Agent applications, supporting hot switching and backup recovery.
- Local intelligent routing : Built-in fuse, health monitoring and failover queue, automatically switches to backup node when the main provider fails, without interrupting the workflow.
- Usage and cost tracking : Real-time statistics on request count, token consumption, cache hits, subscription quota and fees, with support for filtering by application and provider.
- Secure local storage : All configurations and API keys are stored in a local SQLite database, supporting full schema migration, and data never leaves the country.
- MCP / Skills / Prompt Management : Unified management of MCP services, Skills packs, Prompts templates, and Hermes Memory, supporting cross-application synchronization and symbolic link deployment.
- Centralized session management : Aggregates local session records from various agents, supporting search, recovery, and deletion, breaking down session silos between applications.
- Model Mapping and Protocol Conversion : Maps Claude official role IDs or Codex GPT model names to actual models from third-party vendors, automatically completing protocol conversion between the Responses API and the Chat Completions API.
How to use CC Switch
- Download and install : Visit the CC Switch official website https://ccswitch.io/ to download the installation package for your system, and launch CC Switch after installation.
- Select the management target : On the main interface, select the Agent application that needs to be managed and activate the toolbar monitoring.
- Add a supplier : Go to the supplier configuration page, fill in the third-party API endpoint address and authentication token, and the system will automatically check connectivity after saving.
- Enable local routing : For applications such as Codex or Claude Desktop, enable the “Require local route mapping” switch to keep the proxy running.
- Configure model mapping : In the mapping settings, specify the name of the third-party model that is actually requested for the Sonnet/Opus/Haiku or GPT series.
- Synchronize Skills : Import existing skill packs or install them from ZIP on the Skills management page, and synchronize them to the configuration directories of each Agent in real time via symbolic links.
- Monitor usage : View real-time request logs, token consumption and cost curves in the usage statistics panel, and set quota alerts.
CC Switch’s core advantages
-
Full application coverage : The industry’s only unified manager that simultaneously supports Claude dual-platform, Codex, Gemini, OpenCode, OpenClaw, and Hermes.
-
Zero-intrusion integration : No need to modify the Agent source code, you can use third-party models simply by pointing the environment variable to the local agent.
-
High-availability routing : Built-in circuit breaking and failover, switching to a backup provider in milliseconds when the main line fails, ensuring development continuity.
-
Cost transparency : Fine-grained usage tracking helps developers identify high-consumption processes, and in practice, it can save about 30% of API expenses.
-
Open source and free : The MIT license is completely open source, with community-driven iteration and no barriers to commercial licensing.
CC Switch project address
- Project website : https://ccswitch.io/
- GitHub repository : https://github.com/farion1231/cc-switch
Comparison of CC Switch with similar products
| Dimension | CC Switch | LiteLLM | ChatWise |
|---|---|---|---|
| Product Form | Cross-platform desktop application (Tauri) | Python backend/proxy service | Cross-platform desktop client |
| Management Scope | 7 major agent applications + local routing | General-purpose LLM API Gateway | Multi-model dialogue client |
| Protocol Conversion | Automatic Responses ↔ Chat Completions | Supports unified interface for 100+ models | Depends on the official client |
| Failover | Built-in fuse + health monitoring + queue | Supports load balancing and fallback | none |
| Skills/MCP | Unified management and cross-application synchronization | You need to configure it yourself. | Partial support for MCP |
| Usage tracking | Two-dimensional statistics by application × Provider | Enterprise-level billing and observation | Basic Token Statistics |
| Open source license | MIT | MIT | Closed-source commercial software |
| Target users | Multi-Agent Heavy Developer | Enterprise-level LLM Operations and Maintenance | Personal AI conversation with users |
Application scenarios of CC Switch
-
Multi-model A/B testing : Developers can quickly compare the code generation effects of models such as Claude, GPT, DeepSeek, and GLM within the same project.
-
Enterprise-grade API disaster recovery : Automatic switching between primary and backup providers avoids development interruptions caused by rate limiting or failure of a single provider.
-
Refined cost control : Team leaders track token consumption by project or member and optimize model call strategies.
-
Skills Asset Accumulation : Centrally manage custom MCP tools and Prompt templates, and synchronize new members to all Agents with one click.
-
Cross-application session continuation : When switching between Claude Code and Codex, the context session is preserved, reducing the cost of repeated communication.