Using MCP Tools

Connect external tools and services using the Model Context Protocol

The Model Context Protocol (MCP) allows you to connect external tools and services using a standardized protocol, enabling you to integrate APIs and services directly into your workflows. With MCP, you can extend SteelEngine's capabilities by adding custom integrations that work seamlessly with your agents and workflows.

What is MCP?

MCP is an open standard that enables AI assistants to securely connect to external data sources and tools. It provides a standardized way to:

  • Connect to databases, APIs, and file systems
  • Access real-time data from external services
  • Execute custom tools and scripts
  • Maintain secure, controlled access to external resources

Configuring MCP Servers

MCP servers provide collections of tools that your agents can use. Configure them in workspace settings:

  1. Navigate to your workspace settings
  2. Go to the MCP Servers section
  3. Click Add MCP Server
  4. Enter the server configuration details
  5. Save the configuration

You can also configure MCP servers directly from the toolbar in an Agent block for quick setup.

Refresh Tools

Click Refresh on a server to fetch the latest tool schemas and automatically update any agent blocks using those tools with the new parameter definitions.

Using MCP Tools in Agents

Once MCP servers are configured, their tools become available within your agent blocks:

Using MCP Tool in Agent Block
  1. Open an Agent block
  2. In the Tools section, you'll see available MCP tools
  3. Select the tools you want the agent to use
  4. The agent can now access these tools during execution

Standalone MCP Tool Block

For more granular control, you can use the dedicated MCP Tool block to execute specific MCP tools:

Standalone MCP Tool Block

The MCP Tool block allows you to:

  • Execute any configured MCP tool directly
  • Pass specific parameters to the tool
  • Use the tool's output in subsequent workflow steps
  • Chain multiple MCP tools together

When to Use MCP Tool vs Agent

Use Agent with MCP tools when:

  • You want the AI to decide which tools to use
  • You need complex reasoning about when and how to use tools
  • You want natural language interaction with the tools

Use MCP Tool block when:

  • You need deterministic tool execution
  • You want to execute a specific tool with known parameters
  • You're building structured workflows with predictable steps

Permission requirements

MCP access follows the roles assigned to the person in the target workspace. Seeded and custom roles can carry these catalog permissions:

ActionRelevant catalog permissions
View the workspace MCP areaworkspace:read_mcp
List server metadata and available toolsmcp_server:read, mcp_tool:list
Add or update a serverworkspace:manage_mcp, mcp_server:manage
Refresh a server's toolsmcp_server:refresh
Delete a servermcp_server:delete
Call an MCP toolmcp_tool:call, with the applicable MCP scope and workflow permissions

Role names do not determine the result. A custom role with only MCP read permissions can inspect connections without changing or calling them. A role with call permission can use tools only where its assignment scope and the workflow's access also apply.

See Permission Catalog Reference and How Effective Access Is Calculated.

Common Use Cases

Database Integration

Connect to databases to query, insert, or update data within your workflows.

API Integrations

Access external APIs and web services that don't have built-in SteelEngine integrations.

File System Access

Read, write, and manipulate files on local or remote file systems.

Custom Business Logic

Execute custom scripts or tools specific to your organization's needs.

Real-time Data Access

Fetch live data from external systems during workflow execution.

Security considerations

  • SteelEngine policy controls who can configure, inspect, select, and call MCP tools
  • The remote server's credentials determine what the remote service itself can access
  • Always verify MCP server sources before installation
  • Use environment variables for sensitive configuration data
  • Review MCP server capabilities before granting access to agents

Troubleshooting

MCP Server Not Appearing

  • Verify the server configuration is correct
  • Check that you have the required permissions
  • Ensure the MCP server is running and accessible

Tool Execution Failures

  • Verify tool parameters are correctly formatted
  • Check MCP server logs for error messages
  • Ensure required authentication is configured

Permission Errors

  • Open Organization Settings → Access and inspect your effective roles for the workspace
  • Confirm the role includes the catalog permission for the attempted MCP action
  • Check direct, group, inherited, and organization-wide assignment paths
  • Check if the MCP server requires additional authentication
  • Verify the server is properly configured for your workspace

Common Questions

When you add MCP tools to an Agent block, the AI decides which tools to use based on the conversation context and its reasoning. This is best for dynamic, conversational workflows. The standalone MCP Tool block executes a specific tool with explicit parameters every time, giving you deterministic, predictable execution. Use Agent blocks for flexible reasoning and MCP Tool blocks for structured, repeatable steps.
Anyone whose effective roles grant the matching MCP catalog permissions. Reading, managing, refreshing, deleting, listing tools, and calling tools are separate permissions, so a custom role can receive only the actions it needs.
MCP servers are configured per workspace. Each workspace maintains its own set of MCP server connections. If you need the same MCP server in multiple workspaces, you need to configure it separately in each workspace's settings.
Click the Refresh button on the MCP server in your workspace settings. This fetches the latest tool schemas from the server and automatically updates any agent blocks that use those tools with the new parameter definitions.
Yes. Assign an MCP-focused role to a permission group at organization or workspace scope. Members inherit that role through the group. Enterprise feature restrictions can also disable MCP tool availability for a group; both policy permission and feature availability must allow the action.
If the MCP server is unreachable during execution, the tool call will fail and return an error. In an Agent block, the AI may attempt to handle the failure gracefully. In a standalone MCP Tool block, the workflow step will fail. Check MCP server logs and verify the server is running and accessible to troubleshoot connectivity issues.

On this page