C2S Agent is an AI Agent construction platform that also exposes MCP (Multi-Channel Protocol) services. It lets you configure Agents with an HTTP interface and bind them to MCP tools so you can orchestrate multiple MCP services through a single platform. With C2S Agent, you can create configurable MCP tools by defining their HTTP endpoints, attach them to your Agent workflows, and coordinate tool usage across multiple MCP services on a single port. The system supports inter-agent communication (A2A) and MCP-based tool invocation, enabling teams to solve problems by composing services and tools programmatically.

To use the MCP capabilities, first define your MCP services by specifying the HTTP interfaces (endpoints, methods, headers, and body schemas) that your tools expose. Then create and bind these MCP services to an Agent, so the Agent can call these tools during its reasoning process. You can run multiple MCP services on a single port by leveraging the provided MCP protocol server implementation. The platform is designed to work with standard HTTP endpoints, so any service with a REST-like interface can be exposed as an MCP tool.

Prerequisites:

• Python 3.9+ (or compatible Python environment)
• Optional: Docker and Docker Compose for quick local deployment
• PostgreSQL and/or MongoDB if you plan to persist data

Option A: Quick start using Python environment

1. Clone the repository git clone https://github.com/C2SAgent/c2sagent.git cd c2sagent

2. Create and activate a Python virtual environment python3 -m venv .venv source .venv/bin/activate

3. Install dependencies pip install -r requirements.txt

4. Configure environment variables

   • Create a .env file with your DATABASE_URL, MONGO_URI, and other required settings
   • Example content: DATABASE_URL=postgresql://user:pass@localhost:5432/c2sagent MONGO_URI=mongodb://localhost:27017/c2sagent ENV=development

5. Run the MCP server (using uvx as the Python-based runner) uvx c2sagent

   or if you prefer a module-based invocation

   python -m c2sagent.mcp_server

Option B: Quick start using Docker Compose

1. Ensure Docker and Docker Compose are installed
2. From the project root, start backend and frontend services: docker-compose up --build -d
3. Follow the project’s docker/setup instructions to configure environment variables for the containers.
4. Access the MCP interfaces via the exposed ports defined in docker-compose.yml.

Notes:

• The exact module names for running the MCP server may vary depending on how the project is packaged; if the entry point is different, adjust the commands accordingly.
• If you encounter missing dependencies, install or upgrade via pip and ensure your virtual environment is activated.

Tips and common issues:

• Ensure the MCP HTTP tool endpoints you expose are reachable from the MCP server (consider network restrictions and CORS as needed).
• When binding multiple MCP services to a single port, the server’s configuration (mcp_server_id) must distinguish each service; consult the A2A/MCP server code for the correct identifiers.
• Keep PostgreSQL and MongoDB running as required by the backend; ensure credentials and network access are configured in .env or your environment.
• If using Docker, you can leverage docker-compose to spin up both backend and frontend stacks; remember to add your domain or localhost to nginx configurations when needed.
• For debugging, set DEBUG=true and review logs emitted by the MCP server to trace tool invocation and agent interactions.