v0.5.0 tool setup

✨ Release Notes

Tool setup

Users can now select which MCP tools to enable and save the configuration in browser storage.
Frequently used combinations can be saved as presets for quick tool setup switching.

The default app toolkit (e.g., chart) can also be toggled on or off.

All settings are stored locally in the browser, providing a fast and seamless UX.

For example, I use a preset that includes several GitHub tools and one WebSearch tool when I'm exploring recent repositories.

v0.4.1

✨ Release Notes

🐞 Bug Fixes & Stability Improvements

• Chat Error Handler Improvements
  • When an error occurs during chat, a clear error message is displayed and users can continue the conversation seamlessly after the error.
• MCP Manager Single Initialization
  • Refactored to ensure the MCP Client Manager is initialized only once using Next.js’s instrumentation.ts.
• Manual Tool Call Issue Resolved
  • The toolChoice state is now stored per message, decoupling it from the global state for more reliable tool invocation UI.

🆕 Feature Updates

• Temporary Chat Sidebar
  • The temporary chat feature is now provided as a sidebar instead of a separate page.
  • Designed for quick, simple questions that you don’t want to include in the main chat context.
  • Easily toggled with a shortcut key—surprisingly useful in practice!

---

🗓️ What’s Next

• MCP Tool Select UX Improvements
  • A new, more intuitive UX for configuring MCP tools is in the works.
• OpenAI Realtime Audio Chat
  • Real-time audio chat mode integrated with MCP tools is under development.
• Self Hosting
  • Self-hosting support is planned; currently looking for contributors.

v0.4.0

🆕 New Feature: Built-in Chart Visualization

You can now visualize chatbot responses as pie, bar, or line charts using a built-in tool.
Just ask the chatbot to generate a chart based on structured data — no setup required!

New: Authentication

✨ New: Authentication

• Added user authentication logic.
• This is part of the groundwork for self-hosting support, enabling user-based project context and access control in the future.

🐳 Self-Hosting Preparation

• Structural improvements are underway to support self-hosting via Docker and Vercel.
• On Vercel, running a local mcp-server is not possible. Only external remote MCP servers using SSE are supported.
• Local SQLite DB support will be deprecated. PostgreSQL is recommended going forward.

🔄 Database Schema Overhaul

• A complete internal redesign of the database schema has been implemented.
• If you encounter migration errors, the simplest solution is to drop all existing tables and run:

pnpm db:migrate

⚠️ This is a one-time reset to lay the foundation for more stable updates in future releases.

---

🗭 Upcoming Features

• 🛠 Improved Self-Hosting Support
  Docker and Vercel deployments (Vercel supports remote MCP servers only)

• 🗃 MCP Server Config: File → DB
  .mcp-config.json will be removed and replaced with persistent DB-based management.

• 🔄 MCP Workflow Support
  Toolchain-level orchestration powered by the MCP system. (Stay tuned!)

• ❄️ Cold Start Handling for MCP Servers
  Logic to handle delayed startup behavior is coming soon.

v0.2.0 Tool Choice

Tool Usage Control with Tool Choice Mode

• Introduced toolChoice mode to control how tools are invoked during chat
  • auto: model automatically calls tools when needed
  • manual: user confirmation is required before calling any tool
  • none: disables all tool usage
• Added a dropdown UI for switching modes
• Added ⌘P keyboard shortcut to toggle modes quickly

v0.1.0

What's New

Projects to organize related conversations with custom AI instructions
Chat-to-Project Conversion that automatically summarizes existing chats into project instructions

Convert any chat to a project by clicking "Summarize as Project" in the chat menu, or create new projects from the sidebar!

v0.0.4

🚀 MCP Client Chatbot – New Features!

• 💬 @mention support: Quickly trigger MCP tools by typing @ in the chat input.
• ⚙️ Tool status control in input UI: Instantly view and toggle tool activation directly from the chat bar.

v0.0.3

🆕 MCP Client Chatbot – v0.0.3 Release

This update introduces a new testing UI, a smoother chat experience, and expanded model support!

---

✨ What's New

• 🧪 MCP Tool Test Page
  Added a dedicated interface to run MCP tools independently without going through the chatbot. Perfect for debugging and experimenting with tools in isolation.

• 💬 Smoother Chat Message Animation
  Chat messages are now streamed and displayed with a natural fade-in effect, split by ' ' (spaces) for smoother, more dynamic rendering.

• 🤖 Support for OpenAI o4-mini Model
  You can now use OpenAI’s lightweight o4-mini model within your MCP config.

Add GPT-4.1 Support and Fix MCP Configuration Page Scroll Bug

What's New

• Added support for GPT-4.1 model
• Fixed scroll bug in MCP configuration page

Initial Release – MCP Client Chatbot

This is the first public release of MCP Client Chatbot, a flexible and extensible AI chat interface powered by the Model Context Protocol (MCP).

✨ Highlights

• 🔌 Supports multiple AI providers: Easily connect to OpenAI, Anthropic, Google, Ollama, and more.
• ⚙️ File-based configuration: Manage your AI tool chains using a simple JSON config or UI.
• 💬 Streamlined chat experience: Built with Vercel AI SDK and Next.js for seamless streaming chat.
• 🧪 MCP Server Test Page: Try out tools directly, without LLMs, using a standalone testing interface.

Ideal for developers building custom AI workflows or evaluating tools across multiple providers.