Advanced Configuration

Detailed configuration options for haive-mcp and its 1900+ MCP servers.

Configuration Overview

Haive-mcp supports multiple configuration levels:

  1. Automatic - Zero configuration with smart defaults

  2. Category-based - Configure by server categories

  3. Server-specific - Fine-tune individual servers

  4. Custom - Full control over all aspects

Configuration Methods

Environment Variables

Configure MCP behavior through environment variables:

# Set MCP home directory
export MCP_HOME=/opt/mcp

# Configure timeout
export MCP_TIMEOUT=60

# Enable debug mode
export MCP_DEBUG=true

# Set server categories
export MCP_CATEGORIES=core,ai_enhanced

# Configure parallel installations
export MCP_MAX_PARALLEL=5

Configuration Files

Use YAML or JSON configuration files:

# mcp_config.yaml
mcp:
  enabled: true
  auto_discover: true
  auto_install: true
  timeout: 30

  categories:
    - core
    - ai_enhanced
    - time_utilities

  servers:
    filesystem:
      enabled: true
      timeout: 10
      args:
        - "--root-dir=/data"

    github:
      enabled: true
      env:
        GITHUB_TOKEN: "${GITHUB_TOKEN}"

  discovery:
    cache_ttl: 3600
    max_cache_size: 1000

  installation:
    parallel: true
    max_workers: 5
    retry_attempts: 3

Programmatic Configuration

Configure directly in code:

from haive.mcp.config import MCPConfig, MCPServerConfig

# Detailed configuration
config = MCPConfig(
    enabled=True,
    auto_discover=True,
    auto_install=True,
    timeout=30,

    # Server categories
    categories=["core", "ai_enhanced"],

    # Specific server configs
    servers={
        "filesystem": MCPServerConfig(
            transport="stdio",
            command="npx",
            args=["-y", "@modelcontextprotocol/server-filesystem", "/data"],
            env={"NODE_ENV": "production"},
            timeout=10
        ),
        "custom": MCPServerConfig(
            transport="http",
            url="http://localhost:8080",
            headers={"Authorization": "Bearer token"}
        )
    },

    # Discovery settings
    discovery_config={
        "cache_enabled": True,
        "cache_ttl": 3600,
        "parallel_discovery": True
    },

    # Installation settings
    installation_config={
        "parallel": True,
        "max_workers": 5,
        "npm_registry": "https://registry.npmjs.org",
        "pip_index": "https://pypi.org/simple"
    }
)

# Use with agent
agent = EnhancedMCPAgent(
    name="configured",
    engine=AugLLMConfig(),
    mcp_config=config
)

Server Categories Configuration

Core Category

Essential tools for everyday operations:

core_config = {
    "filesystem": {
        "root_dirs": ["/home/user", "/data"],
        "allow_write": True
    },
    "postgres": {
        "connection_string": "postgresql://localhost/mydb",
        "pool_size": 10
    },
    "brave-search": {
        "api_key": "YOUR_API_KEY",
        "safe_search": "moderate"
    }
}

AI Enhanced Category

AI-powered tools configuration:

ai_config = {
    "openai": {
        "api_key": "YOUR_OPENAI_KEY",
        "model": "gpt-4",
        "temperature": 0.7
    },
    "anthropic": {
        "api_key": "YOUR_ANTHROPIC_KEY",
        "model": "claude-3"
    },
    "huggingface": {
        "token": "YOUR_HF_TOKEN",
        "models": ["bert-base", "gpt2"]
    }
}

Transport Configuration

STDIO Transport

For command-line based servers:

stdio_config = MCPServerConfig(
    transport="stdio",
    command="node",
    args=["server.js"],
    cwd="/opt/servers/custom",
    env={
        "NODE_ENV": "production",
        "LOG_LEVEL": "info"
    },
    startup_timeout=10,
    shutdown_timeout=5
)

HTTP Transport

For web-based servers:

http_config = MCPServerConfig(
    transport="http",
    url="https://api.example.com/mcp",
    headers={
        "Authorization": "Bearer token",
        "X-API-Version": "2.0"
    },
    timeout=30,
    retry_config={
        "max_attempts": 3,
        "backoff": "exponential",
        "max_delay": 10
    }
)

SSE Transport

For real-time streaming:

sse_config = MCPServerConfig(
    transport="sse",
    url="https://stream.example.com/events",
    reconnect=True,
    reconnect_delay=5,
    max_reconnect_attempts=10
)

Performance Tuning

Connection Pooling

performance_config = {
    "connection_pool": {
        "min_size": 5,
        "max_size": 20,
        "timeout": 30,
        "idle_timeout": 300,
        "max_lifetime": 3600
    },
    "cache": {
        "enabled": True,
        "type": "redis",
        "redis_url": "redis://localhost:6379",
        "ttl": 3600,
        "max_entries": 10000
    },
    "batch_processing": {
        "enabled": True,
        "batch_size": 100,
        "flush_interval": 5
    }
}

Resource Limits

limits_config = {
    "memory": {
        "max_heap": "512M",
        "max_rss": "1G"
    },
    "cpu": {
        "max_usage": 0.8,
        "nice": 10
    },
    "io": {
        "max_open_files": 1000,
        "max_disk_usage": "10G"
    },
    "network": {
        "max_connections": 100,
        "bandwidth_limit": "10M"
    }
}

Security Configuration

Authentication

auth_config = {
    "authentication": {
        "type": "oauth2",
        "client_id": "YOUR_CLIENT_ID",
        "client_secret": "YOUR_SECRET",
        "token_url": "https://auth.example.com/token",
        "scopes": ["read", "write"]
    },
    "api_keys": {
        "openai": "${OPENAI_API_KEY}",
        "anthropic": "${ANTHROPIC_API_KEY}",
        "github": "${GITHUB_TOKEN}"
    }
}

Encryption

encryption_config = {
    "tls": {
        "enabled": True,
        "verify": True,
        "cert_file": "/path/to/cert.pem",
        "key_file": "/path/to/key.pem",
        "ca_bundle": "/path/to/ca.pem"
    },
    "secrets": {
        "provider": "vault",
        "vault_url": "https://vault.example.com",
        "vault_token": "${VAULT_TOKEN}"
    }
}

Logging Configuration

logging_config = {
    "level": "INFO",
    "format": "json",
    "outputs": [
        {
            "type": "file",
            "path": "/var/log/mcp/agent.log",
            "rotation": "daily",
            "retention": 7
        },
        {
            "type": "console",
            "format": "colored"
        },
        {
            "type": "syslog",
            "host": "localhost",
            "port": 514
        }
    ],
    "filters": {
        "exclude_sensitive": True,
        "redact_patterns": ["api_key", "token", "password"]
    }
}

Deployment Configurations

Development

dev_config = MCPConfig(
    enabled=True,
    auto_discover=True,
    auto_install=True,
    categories=["core"],
    timeout=60,
    debug=True,
    verbose=True
)

Production

prod_config = MCPConfig(
    enabled=True,
    auto_discover=False,  # Use explicit servers
    auto_install=False,   # Pre-installed servers only
    servers={
        # Explicitly configured servers
    },
    timeout=30,
    retry_attempts=3,
    health_check_interval=60,
    monitoring_enabled=True,
    metrics_port=9090
)

Configuration Validation

from haive.mcp.config import validate_config

# Validate configuration
errors = validate_config(config)
if errors:
    for error in errors:
        print(f"Configuration error: {error}")
else:
    print("Configuration valid")

Next Steps