What is MCP?
Model Context Protocol (MCP) is an open standard that lets AI assistants interact with external tools and data sources. Ekso’s MCP server exposes 104 tools that AI agents can use to read and write project data — create items, plan sprints, manage documents, query finance reports, wire up webhook-style rule integrations, and more. MCP is supported by Claude, Cursor, VS Code, Windsurf, and other AI-powered tools.Server endpoint
The MCP server is available at:Connecting
Claude Desktop
Add this to your Claude Desktop configuration (claude_desktop_config.json):
Claude Code
Add via the CLI:Cursor
Add this to your Cursor MCP settings (.cursor/mcp.json):
Authentication
MCP requests use the same Bearer token authentication as the REST API. The token is included automatically when you authenticate through your AI tool’s OAuth flow. For tools that support it, the MCP server will initiate OAuth 2.0 authorization using the same OAuth endpoints as the REST API.Tool categories
Ekso’s MCP tools are organized by domain:Resources
In addition to tools, Ekso’s MCP server exposes 10 read-only resources that provide AI-optimized domain knowledge — covering concepts, terminology, and workflows. Resources help agents build context before invoking tools. See the Resources page for the full list.Prompts
Ekso’s MCP server provides 7 conversation prompts — pre-built templates that guide AI agents through multi-step planning workflows like sprint planning, review, and standup. Prompts pre-load live data (boards, cycles, items) so the AI can act immediately. See the Prompts page for details.Deep links
Item, container, and board tool responses include aurl field that links directly to the entity in the Ekso web app. AI agents can present these as clickable links so users can jump straight from an AI response into the application.
Read vs. write tools
Each tool is marked as either read-only or read-write:- Read-only tools query data without side effects. AI agents can call these freely for context gathering.
- Read-write tools create or modify data. These require appropriate permissions and most AI agents will confirm before executing them.
Rate limits
MCP tools are rate-limited separately from the REST API using theMcp rate limiting policy.