Best for
- Users who want to manage MCP servers from a desktop application
- Users connecting ChatGPT, Claude, Qwen, or self-hosted LLM backends
- Developers exploring MCP integrations and testing applications
T
TUUI
Community
Not a standard server
A local AI playground for MCP
This is not a typical connect-and-use MCP server; read the source repository before using it.
- Category
- Dev Tools #219 of 438
- Stars
- ★ 1.2k Very popular
- Runtime
- Prebuilt binary
- Credentials
- Optional API key
- License
- Apache-2.0
- Last commit
53FMRS · D
Based on the README, TUUI is an open-source desktop MCP client maintained by AI-QL, not an MCP server. The repository does not provide a server manifest or tool identifiers for TUUI itself.
Strongest · Documentation 13/20 Weakest · Reliability 8/20
Why each score
- Reliability 8/20
- The README describes a desktop MCP client rather than a standalone MCP server and claims support for tools, prompts, resources, sampling, and elicitation. However, no cached server manifest, CI workflows, committed tests, or verifiable evidence of the initialization handshake and tool behavior is provided. The documentation also reports ENOENT issues, an unresolved macOS runtime issue, and a 90-second timeout, so the happy path is plausible but error and edge-case control is weak.
- Security and permissions 8/20
- The documentation shows support for multiple external LLM endpoints and remote MCP servers through mcp-remote with OAuth. Configurations contain apiKey fields and are stored in localStorage, but there is no evidence of secret protection, data-boundary disclosure, least-privilege controls, tool-call confirmation, or safeguards for dangerous side effects. No explicit malware, credential theft, covert exfiltration, or irreversible default behavior is shown, so no red line is triggered; security evidence remains incomplete.
- Maintenance 12/20
- The project is not archived, declares an Apache-2.0 license, and provides contribution guidance, a code of conduct, and a security-reporting channel. However, the prompt supplies no verifiable release cadence, commit activity, dependency-update evidence, or issue-response history beyond counts. This supports basic maintenance and governance, but not a sustained high-maintenance rating.
- Documentation 13/20
- The README covers platforms, runtime prerequisites, LLM and MCP configuration, remote connections, OAuth-related troubleshooting, startup errors, and timeouts, with several configuration examples. Points are deducted because no cached manifest, server tool parameters and limits, cost information, or clear privacy/data-flow boundaries are provided; it also includes placeholder URLs, a platform typo, and reliance on external documentation.
- Setup experience 12/20
- The README provides download paths for Windows, Linux, and macOS, lists Node.js, Python/UV, and Docker prerequisites, and includes MCP configuration examples. However, setup depends on manual configuration, runtime installation, path and permission changes, and platform-specific troubleshooting. The remote example uses a placeholder domain, and there is no reproducible evidence from installation through a verified working connection, making the setup usable but fragile.
Static review · not runListed 2026-07-30
Read the FMRS scoring method →Fit and risk
What it can accessRuns commands or codeUses the network
Not for
- Users seeking an MCP server callable by another MCP client
- Users who do not want to configure an LLM backend or run local MCP server processes
- Users seeking a server directory with a predefined tool list
Required permissions
- Access to configured LLM APIs or self-hosted backends
- Permission to launch configured local commands and runtimes for MCP servers
- Network access to configured remote MCP servers and related services
- Access to read and save application configuration and local storage
Risks and side effects
- Configured MCP servers are launched and executed according to their settings, so behavior depends on those servers and commands
- Configured LLM APIs may process chat content, tool-call content, and API keys
- Remote MCP connections and OAuth configuration may be affected by network, browser-cache, or callback issues
- Initial installation or updates of npx, uv, or uvx runtimes may cause connection timeouts
Setup
Before you start
Runtime：Prebuilt binary
Download and run the application for Windows, Linux, or macOS from GitHub Releases. Developers can build the project using the English or Chinese installation guide in the repository. Before using MCP features, configure an LLM backend that supports tool calling; depending on the MCP server type, Node.js, Python with UV, or Docker may also be required.
Check that it works
Download and run TUUI from the Releases page, add an MCP server (e.g. the cloudflare mcp-remote example) to mcp.; if the server's tools appear in the client's tool list after startup, the connection works.
Troubleshooting
- Verify that commands such as uv, uvx, and npx run correctly on the current system
- For ENOENT errors, run the MCP server locally and configure its absolute path
- On macOS with NVM, check MCP command paths and permissions
- For initialization timeouts beyond 90 seconds, manually install the relevant pip or npm dependencies first
- For remote authentication failures, check OAuth auto-redirect behavior and clear the authentication-page cache
Things to try
Once connected, you can ask your AI assistant things like:
- Search the MCP Registry and list available MCP servers
- Call a tool from one of the configured MCP servers and show the call trace
- Read a resource from a loaded MCP server and list its available prompts
Use cases
Connecting to and using MCP server tools
Configuring and orchestrating multiple LLM APIs
Tracing MCP tool calls
Exploring MCP primitives, elicitation, and MCP Bundles
Overview
TUUI is a desktop MCP client and local LLM chat application for integrating AI tools through the Model Context Protocol and dynamically orchestrating cross-vendor LLM APIs. It supports MCP tools, prompts, resources, sampling, elicitation, registry discovery, and MCP Bundles (.mcpb).
Similar servers
Context7
80 · B
Upstash's official server providing up-to-date third-party library docs for AI coding assistants
NPM Sentinel MCP Server
79 · B
AI-powered NPM package analysis MCP server
MartinLoop MCP Server
79 · B
An independent control layer that gives AI coding agents budgets, brakes, and receipts
GitHub MCP Server
76 · B
GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language
Source revision 0be513448dc6 Data synced 2026-10-19 Read the FMRS scoring method