MCP Servers
Connect to Model Context Protocol servers to access external tools and data sources.
Model Context Protocol (MCP) Servers are interfaces that expose data and functionality from external systems or data sources. While Axellero Studio is a design-time tool for building integrations, the AI agents and workflows you create become MCP clients at runtime, connecting to MCP servers to retrieve data and execute tasks.
What are MCP Servers?
MCP (Model Context Protocol) Servers are standardized interfaces that:
- Expose external tools and data sources to LLM applications like Axellero
- Facilitate communication between AI systems and external systems
- Provide structured access to databases, file systems, APIs, and services
- Enable tool integration beyond built-in connector capabilities
Runtime MCP Integration
When you build AI agents and workflows in Axellero Studio, they become MCP Clients at runtime:
- Agents and workflows connect to MCP-compliant servers during execution
- Retrieve data from external systems via MCP protocol
- Execute tasks and operations through MCP servers
- Integrate with tools that provide MCP interfaces
Axellero Studio provides the design-time interface for configuring these MCP connections.
Common MCP Server Types
System Access Servers
Provide access to system resources and tools:
| Server Type | Examples | Use Cases |
|---|---|---|
| File System | Local file operations, cloud storage | Read/write files, directory operations, file management |
| Database | SQL databases, NoSQL stores | Query execution, data retrieval, database operations |
| Web Search | Search engines, web crawlers | Information gathering, research, content discovery |
| Command Line | Shell access, system commands | System administration, script execution, automation |
Development Tools Servers
Integrate with development and productivity tools:
| Server Type | Examples | Use Cases |
|---|---|---|
| Version Control | Git repositories, SVN | Code management, version tracking, collaboration |
| Issue Tracking | Jira, GitHub Issues, Linear | Bug tracking, feature requests, project management |
| Documentation | Wikis, knowledge bases | Information retrieval, content management |
| Communication | Slack, Teams, email systems | Messaging, notifications, team coordination |
Business System Servers
Connect to enterprise and business applications:
| Server Type | Examples | Use Cases |
|---|---|---|
| CRM Systems | Salesforce, HubSpot | Customer data, sales pipeline, contact management |
| ERP Systems | SAP, Oracle ERP | Business processes, resource planning, operations |
| Analytics | Business intelligence tools | Data analysis, reporting, insights |
| Custom APIs | Internal services, legacy systems | Proprietary integrations, specialized workflows |
Adding MCP Servers
Creating a New MCP Server Connection
- Navigate to Integrations: Go to the Integrations section
- Select MCP Servers Tab: Click on the "MCP Servers" tab
- Add New Server: Click the "+" button to create a new connection
- Configure Server Details: Enter server information and settings
- Test Connection: Verify the server is accessible and responding
- Save Configuration: Store the MCP server setup
Configuration Options
| Setting | Description | Implementation | Required |
|---|---|---|---|
| Name | Descriptive server identifier | Free text field | Yes |
| Description | Purpose and capabilities | Free text field | No |
| Server URL | MCP server endpoint | Full URL with transport type | Yes |
| Transport Type | Connection method | SSE or Streamable HTTP | Yes |
Authentication Methods:
| Method | Configuration | Implementation | Use Cases |
|---|---|---|---|
| OAuth 2.0 | Client ID, Client Secret, Auth URL, Token URL | Full OAuth flow with PKCE | Secure third-party services |
| Custom Headers | Header name/value pairs | HTTP headers in requests | API keys, bearer tokens |
| No Authentication | None | Direct connection | Public or internal servers |
Using MCP Servers
Tool Integration
MCP tools automatically become available as workflow nodes:
| Integration Type | How It Works | Capabilities |
|---|---|---|
| AgentFlow | MCP tools appear in node palette | Agents can call external tools during conversations |
| Low-Code Workflows | Automatic node generation from tool schemas | Visual workflow building with MCP tool nodes |
| Dynamic Discovery | Tools sync automatically when server connects | New tools become available without reconfiguration |
MCP Server Status
Server States
| State | Description | Visual Indicator | Actions Available |
|---|---|---|---|
| Connected | Server responding and tools synced | Green indicator | View tools, disconnect, edit |
| Auth Required | OAuth flow needed | Orange indicator | Complete authentication |
| Error | Connection or authentication failed | Red indicator | View error, retry, edit configuration |
| Pending | Connection attempt in progress | Loading indicator | Wait or cancel |
Implementation Details
Transport Types
| Transport | Protocol | Use Cases | Configuration |
|---|---|---|---|
| SSE | Server-Sent Events | Real-time updates, streaming responses | WebSocket-like connection |
| Streamable HTTP | HTTP with streaming | Standard REST-like interface | Traditional HTTP requests |
Tool Discovery
When an MCP server connects:
- Automatic Discovery: Server tools are automatically detected
- Schema Validation: Tool input schemas are validated using JSON Schema
- Node Generation: Tools become available as workflow nodes
- Real-time Updates: New tools appear immediately when server updates
Execution Flow
| Step | Process | Timeout | Error Handling |
|---|---|---|---|
| Tool Call | Agent/workflow invokes MCP tool | 60 seconds | Automatic retry with exponential backoff |
| Parameter Validation | Input validated against JSON Schema | Immediate | Validation errors returned to caller |
| Server Execution | MCP server processes tool request | Server-dependent | Server errors propagated to workflow |
| Result Return | Tool output returned to agent/workflow | Immediate | Structured response or error details |