HyperChat

HyperChat is a Chat client and local AI Agent platform that pursues openness. It provides a chat experience by connecting to the APIs of multiple LLMs, and enables productivity tool integration through the MCP protocol. Its core philosophy is "AI as Code": let every project have its own dedicated AI brain, drive AI capabilities through configuration files, and make capabilities localized, portable, and manageable under version control.

1. Main Features

• Fully localized: data never leaves the local environment, ensuring controllable privacy and security.
• Agent memory: the AI Agent can understand and remember project context.
• Configuration as code: all AI capabilities are configured through files and support Git management.
• Deep tool integration: supported through the MCP protocol, it can directly operate on the local file system.
• One-click migration: a complete .hyperchat/ configuration directory migrates with the project.
• Multiple @ file paths: intelligently recognizes multiple file references and supports complex file operations.
• Agent custom commands: a shortcut command system driven by Markdown templates.

1. Dual-Mode Architecture HyperChat 2.0 adopts a dual-frontend architecture with Web and CLI, optimizing the experience for different scenarios.

2. Web frontend: multi-workspace collaboration center The design philosophy is project-level collaboration and unified resource management. It features multi-workspace tab management (multiple project workspaces can be opened at the same time and switched with one click), workspace-level Agent collections and MCP service pools, graphical configuration, real-time monitoring, and data statistics. Applicable scenarios include project development, team collaboration, workspace management, and visual operations.

3. CLI frontend: Agent-first rapid interaction The design philosophy is Agent-centric, enabling extremely fast startup of a single intelligent agent. Agents can be selected directly, and MCP tools can be loaded on demand from the workspace; it includes Agent-specific memory, context, and chat history; the Agent has built-in MCP tools and can fall back to workspace shared resources. Applicable scenarios include quick conversations, automation scripts, command-line workflows, and CI/CD integration.

4. Configuration and Environment Variables HyperChat 2.0 implements a 5-level priority environment variable system: default values (default configuration built into the code), process.env (system environment variables), global .env (~/Documents/HyperChat/.env), workspace .env (the .env file in the project directory), CLI parameters (passedin from the command line, with the highest priority).

5. Underlying Design HyperChat 2.0 adopts a two-layer architecture design: WorkspaceManager-enhanced supports caching of multiple workspace instances; the workspace-level MCP service pool provides unified MCP client management; SSE streaming push enables real-time data synchronization and independent state management for multiple tabs; the configuration merge mechanism intelligently merges global configuration and workspace configuration. On the CLI side, workspace initialization is omitted and Agent instances are accessed directly. The toolchain progressively loads from the Agent's built-in tools to workspace shared resources, while retaining the Agent's independent memory, chat history, and configuration management. Overall, Web mode is workspace-centered and suitable for project-level management and team collaboration; CLI mode is Agent-centered and supports cross-project personal tool use; the two achieve intelligent merging of global and workspace configurations through 5-level priority configuration, while maintaining complete isolation between project-level and personal-level data.

6. Development Status HyperChat 1.0 is a completely hand-written project and is currently being migrated to 2.0. HyperChat 2.0 adopts a collaborative development approach and welcomes development together using AI tools such as Claude Code and GitHub Copilot. The project's official website and source code are located at https://github.com/BigSweetPotatoStudio/HyperChat。关于具体安装命令、支持的 Information such as the LLM list, charging methods, and prices is not currently provided in the available materials.