Best for
- Users who want a self-hosted or open-source chat client
- Users who need cross-platform access to AI models
- Developers or enterprises wanting a customizable AI chat UI
N
- Category
- Other #220 of 230
- Stars
- ★ 88.9k Very popular
- Transport
- stdio (local process)
- Runtime
- Node.js 18+ · Docker 20+
- Credentials
- API key / credential required
- License
- MIT
- Last commit
29FMRS · D
NextChat is a flexible open-source AI chat client that supports multiple models and cross-platform deployment. However, it is not an official MCP server, and users must pay attention to security configuration.
Strongest · Maintenance 16/20 Weakest · Reliability 2/20
Why each score
- Reliability 2/20
- The evidence shows a chat web app. The only MCP-related content in the README is 'NextChat Support MCP' and an ENABLE_MCP=true environment variable. There is no MCP server manifest, no tool list, no endpoint, and no protocol implementation in the supplied material; the SERVER MANIFEST is not cached. An MCP init and tool-list handshake cannot be verified, and no CI/tests for key paths are present, so reliability is scored very low.
- Security and permissions 5/20
- No MCP server credential or permission model is exposed. The app uses environment variables and browser-local storage for API keys, and all examples use placeholders. MCP support is toggled by a single environment variable with no documented least-privilege boundaries, confirmation flows, side-effect scoping, or data-flow disclosure. No red-line issue was found, but the MCP security posture is opaque, so the score is 5.
- Maintenance 16/20
- The evidence shows an MIT license, a non-archived repository, an active release line (e.g., v2.15.8), update announcements, roadmap items, and heavy contributor activity. Deductions are made because no dedicated security-response channel is visible in the supplied material and code-level governance cannot be verified from the README alone. Star count was not counted as a scoring factor.
- Documentation 3/20
- The README is detailed for the chat application itself, covering environment variables, Docker, Vercel, and FAQ. However, for MCP server use it only says to set ENABLE_MCP=true and includes a screenshot. There are no MCP server installation, authentication, tool-parameter, limits, cost, or troubleshooting docs, and no manifest evidence is available, so the documentation score is heavily reduced.
- Setup experience 3/20
- For the application there is a Docker one-liner and Vercel deployment flow, but for an MCP server there is no client configuration example, no MCP server install path, and no disclosed connection URL or transport mechanism. The static calibration also prevents a score above 15 without verifiable CI/test evidence, and none is present, making setup very weak.
Static review · not runListed 2026-08-07
Read the FMRS scoring method →Fit and risk
What it can accessUses the network
Not for
- Production-grade AI services requiring official support
- Non-technical users looking for a ready-to-use AI service (may require configuration)
- Enterprises with strict data governance requirements (data is stored locally in the browser by default)
Required permissions
- Network access to call AI model APIs
- Local storage in the browser (e.g., chat history, settings)
- When using MCP, may connect to local or remote tools
Risks and side effects
- This server is not an official product but a community-maintained open-source project, which may be less stable than official versions.
- If self-hosted, API key security is crucial and must be protected.
- Relying on third-party deployment services (e.g., Vercel) may introduce data security risks.
- Enabling MCP features increases the attack surface.
Setup
Before you start
Runtime：Node.js 18+ · Docker 20+
OPENAI_API_KEY
OpenAI API key from platform.openai.com; documented as required.
CODE
Access password(s), comma-separated; optional.
AZURE_API_KEY
Azure API key.
GOOGLE_API_KEY
Google Gemini Pro API key.
ANTHROPIC_API_KEY
Anthropic Claude API key.
BAIDU_API_KEY
Baidu API key.
BAIDU_SECRET_KEY
Baidu secret key.
BYTEDANCE_API_KEY
ByteDance (Doubao) API key.
ALIBABA_API_KEY
Alibaba Cloud API key.
IFLYTEK_API_KEY
iFlytek API key.
IFLYTEK_API_SECRET
iFlytek API secret.
CHATGLM_API_KEY
ChatGLM API key.
DEEPSEEK_API_KEY
DeepSeek API key.
STABILITY_API_KEY
Stability API key.
SILICONFLOW_API_KEY
SiliconFlow API key.
AI302_API_KEY
302.AI API key.
Other optional settings (27)
ENABLE_MCP
Set to true to enable the MCP (Model Context Protocol) feature; set before build/deploy.
BASE_URL
Overrides the OpenAI API base URL; defaults to https://api.openai.com.
OPENAI_ORG_ID
OpenAI organization ID; optional.
AZURE_URL
Azure OpenAI deploy URL, e.g. https://{resource}/openai.
AZURE_API_VERSION
Azure API version string.
GOOGLE_URL
Google Gemini Pro API URL.
ANTHROPIC_API_VERSION
Anthropic Claude API version.
ANTHROPIC_URL
Anthropic Claude API URL.
BAIDU_URL
Baidu API URL.
BYTEDANCE_URL
ByteDance API URL.
ALIBABA_URL
Alibaba Cloud API URL.
IFLYTEK_URL
iFlytek API URL.
CHATGLM_URL
ChatGLM API URL.
DEEPSEEK_URL
DeepSeek API URL.
HIDE_USER_API_KEY
Set to 1 to prevent users from entering their own API key.
DISABLE_GPT4
Set to 1 to prevent users from using GPT-4.
ENABLE_BALANCE_QUERY
Set to 1 to let users query balance.
DISABLE_FAST_LINK
Set to 1 to disable parsing settings from URL.
CUSTOM_MODELS
Customize the model list with +/- to add/remove, aliases, and Azure/ByteDance deployment names.
DEFAULT_MODEL
Change the default model.
VISION_MODELS
Add vision capabilities to extra models, comma-separated.
WHITE_WEBDAV_ENDPOINTS
Comma-separated whitelist of full WebDAV endpoints.
DEFAULT_INPUT_TEMPLATE
Customize the default user input preprocessing template.
STABILITY_URL
Custom Stability API URL.
SILICONFLOW_URL
SiliconFlow API URL.
AI302_URL
302.AI API URL.
PROXY_URL
Proxy URL, optionally set for Docker deployment.
- Get an OpenAI API key (optional).
- Deploy with one click on Vercel or run locally (requires Node.js >= 18).
- To enable MCP, set the environment variable ENABLE_MCP=true and then start the application.
Check that it works
Set ENABLE_MCP=true before building and start the app; confirm the MCP plugin panel appears in the chat UI and that plugins like network search or the calculator return results.
Troubleshooting
- If API calls fail, check BASE_URL and API key configuration.
- If the page is inaccessible, check the environment variable CODE (acts as an access password).
- If MCP is not effective, confirm ENABLE_MCP=true is set and restart the application.
Things to try
Once connected, you can ask your AI assistant things like:
- With MCP enabled, search the web for today's tech news
- Use the calculator plugin to compute 128*47
- Use the enabled MCP plugin to look up the current exchange rate
Use cases
Chatting via a web or desktop interface
Managing multiple AI providers (e.g., OpenAI, Claude, Gemini)
Deploying a private AI assistant locally
Overview
NextChat is a cross-platform AI chat client that runs on Web, iOS, macOS, Android, Linux, and Windows. It supports multiple AI model providers, including OpenAI GPT-4, Anthropic Claude, Google Gemini Pro, DeepSeek, and more. NextChat prioritizes privacy by storing all data locally in the browser. MCP functionality can be enabled via the environment variable ENABLE_MCP=true.
Similar servers
Hevy MCP
77 · B
Manage your Hevy workout data from AI assistants via MCP.
Firecrawl MCP Server
75 · B
Firecrawl's official MCP server for web search, scraping, and structured extraction for AI agents
BioMCP
75 · B
One binary. One grammar. Evidence from the biomedical sources you already trust.
Compartment
75 · B
Encrypted, fully offline long-term memory for AI agents
Source revision 9b4ac8e1c543 Data synced 2026-10-19 Read the FMRS scoring method