Praison AI

MCP Lifecycle CLI

CLI commands for MCP server lifecycle and agent integration

Commands for running PraisonAI’s MCP server, inspecting built-in tools, and attaching external MCP servers to agent runs.

Quick Start

bash
# Start the built-in MCP server (stdio)
praisonai mcp serve

# List built-in MCP tools
praisonai mcp list-tools

# Run an agent with an external MCP server
praisonai "What time is it?" --mcp "uvx mcp-server-time"

MCP server commands

Serve

Start the PraisonAI MCP server:

bash
praisonai mcp serve --transport stdio
praisonai mcp serve --transport http-stream --host 127.0.0.1 --port 8080

Common flags: --name, --debug, --response-mode batch|stream, --session-ttl, --log-level, --json.

Inspect built-in tools

bash
# List all tools
praisonai mcp list-tools

# Search tools
praisonai mcp tools search "web"

# Tool details and schema
praisonai mcp tools info tavily_search
praisonai mcp tools schema tavily_search

Health check

Use the doctor command to verify MCP connectivity (not praisonai mcp doctor):

bash
praisonai doctor mcp
praisonai doctor mcp --deep

Using MCP with agents

Basic usage

bash
praisonai "What time is it?" --mcp "uvx mcp-server-time"

Multiple MCP servers

bash
praisonai "Get time and search" \
  --mcp "uvx mcp-server-time" \
  --mcp "npx -y @modelcontextprotocol/server-brave-search"

With environment variables

bash
export BRAVE_API_KEY="your-key"
praisonai "Search for Python" \
  --mcp "npx -y @modelcontextprotocol/server-brave-search" \
  --mcp-env "BRAVE_API_KEY=${BRAVE_API_KEY}"

Connection types

TransportExample --mcp value
Stdio (command)"uvx mcp-server-time"
SSE URL"http://localhost:8080/sse"
HTTP stream"http://localhost:8080"
WebSocket"ws://localhost:8080"

Lifecycle management

Connections opened via --mcp are cleaned up automatically when the agent run finishes:

bash
praisonai "What time is it?" --mcp "uvx mcp-server-time"

Debugging

Verbose agent run

bash
praisonai "Task" --mcp "uvx mcp-server-time" --verbose

Python connectivity check

bash
python -c "
from praisonaiagents import MCP

with MCP('uvx mcp-server-time') as mcp:
    print('Connection: OK')
    tools = mcp.get_tools()
    print(f'Tools: {len(tools)}')
print('Cleanup: OK')
"

Related

MCP CLI Reference

Full MCP command reference

 Doctor CLI

MCP health checks via doctor mcp

 MCP Tool Search

Search and inspect MCP tools

 MCP Lifecycle

SDK-level MCP lifecycle