Klavis AI: Model Context Protocol (MCP) Integration Tool for AI Applications

General Introduction

Klavis AI is an open source platform focused on simplifying the use and integration of the Model Context Protocol (MCP), an open standard that allows AI applications to dynamically connect with external tools and data sources.Klavis AI offers Slack, Discord clients, hosted MCP The server and simple web interface lower the technical threshold, making it easy for non-technical users to use and for developers to quickly build and expand MCP applications. The platform supports the integration of multiple tools and emphasizes security by providing OAuth authentication to ensure secure data access.

Function List

• Provides MCP clients for Slack and Discord, supporting the use of MCP functionality directly from the chat platform.
• Hosting of multiple MCP servers such as ReportGen, Resend mail delivery, Firecrawl deep research, etc.
• Support for creating and managing system MCP server instances via RESTful APIs.
• Provides OAuth authentication to ensure secure integration of tools and data sources.
• Supports Docker and local deployment for developers running MCP servers.
• Includes a variety of MCP features such as document and Markdown conversion, YouTube tools, and more.
• Provides a test and evaluation platform to compare the performance of different MCP servers.

Using Help

Installation and Deployment

The core code for Klavis AI is hosted on GitHub, and developers can clone the repository for local deployment or run the MCP server using Docker. Here are the exact steps:

1. clone warehouse::

   git clone https://github.com/Klavis-AI/klavis.git
   cd klavis

2. Deploying with Docker (recommended)::
   • Make sure Docker is installed.
   • Navigate to the project root directory and locate the target MCP server (e.g. mcp_servers/resend).
   • Build a Docker image, for example:

     docker build -t resend-mcp -f mcp_servers/resend/Dockerfile .
     

   • Run the container:

     docker run -p 5000:5000 resend-mcp
     

   • The server will be running at http://localhost:5000.
3. Local deployment (Node.js or Python)::
   • Make sure Node.js (18.0.0+) or Python (3.12+) is installed.
   • Go to the target MCP server directory, such as mcp_servers/markitdown.
   • Install the dependencies:

     npm install  # 对于 Node.js 服务器
     pip install -r requirements.txt  # 对于 Python 服务器
     

   • Configure environment variables (such as API keys), set in the .env file.
   • Start the server:

     node index.js  # 或 python app.py
     

4. Getting the API key::
   • Visit https://www.klavis.ai to register for an account.
   • Generate an API key in the account settings for authentication.

Using the MCP Client

Klavis AI offers Slack and Discord clients that allow users to utilize MCP features without coding.

• Slack Client::
  • Add the Klavis AI app to your Slack workspace (via the install link provided at https://www.klavis.ai).
  • Use a command such as /klavis reportgen to invoke the ReportGen server to generate reports.
  • Enter a web page URL or search term and the server automatically crawls the data, generates JavaScript code and renders a report.
• Discord Client::
  • Join Klavis AI's Discord server (link at https://www.klavis.ai).
  • Use similar commands, such as !mcp send-email, to send mail or perform other tasks.
  • The client supports interactive prompts to guide the user in entering the necessary parameters.

Featured Function Operation

1. ReportGen Server::
   • Used to generate dynamic reports. The user provides a URL or keywords and the server generates reports via web crawling (Firecrawl) and LLM prompts.
   • Operational Steps:
     • Type /klavis reportgen in Slack.
     • The server returns a link to the report with visual charts and formatted content.
     • Developers can view the open source code (mcp_servers/report_generation) to customize the report template.
2. Resend mail delivery::
   • Integration with the Resend service, support for sending plain text or HTML emails, setting up scheduled delivery, cc'ing and bcc'ing.
   • Configuration Steps:
     • Get the Resend API key and verify the sending mailbox.
     • Set RESEND_API_KEY in the .env file.
     • In Discord type !mcp send-email to:recipient@example.com subject:Test body:Hello.
     • The server processes the request and returns the send status.
3. Firecrawl In-depth Research::
   • Support for LLM clients (e.g., Claude) to perform deep research on web pages and extract structured data.
   • Operational Steps:
     • Start the Firecrawl server (refer to Docker Deployment).
     • The research topic is entered on the client side and the server crawls the web page and returns the analyzed results.
     • Configurable retry logic (FIRECRAWL_RETRY_MAX_ATTEMPTS) optimizes performance.

API Usage

Klavis AI provides a RESTful API through which developers can manage system MCP instances. Common requests include:

• Creating an MCP Server::

  curl --request POST \
  --url https://api.klavis.ai/mcp-server/instance/create \
  --header 'Authorization: Bearer <KLAVIS_API_KEY>' \
  --header 'Content-Type: application/json' \
  --data '{ "serverName": "<MCP_SERVER_NAME>", "userId": "<USER_ID>", "platformName": "<PLATFORM_NAME>" }'
  

  Returns the server URL for subsequent operations.

• Setting the Authentication Token::

  curl --request POST \
  --url https://api.klavis.ai/mcp-server/instance/set-auth-token \
  --header 'Authorization: Bearer <KLAVIS_API_KEY>' \
  --header 'Content-Type: application/json' \
  --data '{ "instanceId": "<string>", "authToken": "<string>" }'
  

The API uses HTTPS protocol to ensure data security. Developers can refer to https://docs.klavis.ai for detailed documentation.

caveat

• Ensure that .env files are configured correctly to avoid leaking API keys.
• Docker deployments need to check the port mapping (e.g. 5000:5000).
• Community support is available via Discord (https://discord.gg/3uqNS3KRP2), and questions or code contributions are welcome.

application scenario

1. Automated Reporting in Teamwork
   Marketing teams use ReportGen Server to generate market analysis reports by entering keywords via Slack, saving time on manual organization.
2. AI-driven email workflows
   The customer service team uses Resend Server at Discord to automate customer confirmation emails with timed and multi-recipient support.
3. Developer Tools Integration
   Developers integrate MCP functionality into existing applications via APIs and dynamically call Firecrawl for real-time data analysis.
4. Education and Research
   Using the Firecrawl server, the researchers entered academic topics, quickly fetched web page data and generated structured summaries.

QA

1. What platforms does Klavis AI support?
   Slack, Discord, and the web interface are currently supported, with possible expansion to other platforms in the future.
2. How to get Firecrawl API key?
   Register Firecrawl account (https://firecrawl.dev), generate key in settings, configure to .env file.
3. Is the MCP server free?
   Klavis AI's code and client are open source and free, some hosting servers may involve cloud service fees.
4. How do I contribute code?
   Visit https://github.com/Klavis-AI/klavis to read the contribution guidelines and submit a Pull Request.