MCP Server
The Attune AI MCP Server is a Model Context Protocol implementation that exposes AI workflows, help documentation, memory systems, and utility functions as structured tools for AI clients.
Server architecture
The server uses a modular design where handler mixins provide different categories of functionality to the main EmpathyMCPServer class:
EmpathyMCPServer— Core MCP protocol implementation that coordinates prompts, tools, and resourcesMemoryHandlersMixin— Cross-session memory storage and retrieval for decisions, patterns, and troubleshooting findingsWorkflowHandlersMixin— Execution interface for Attune AI automation workflowsRateLimiter— Sliding-window protection against excessive tool calls
Tool categories
The server exposes five distinct tool groups through MCP:
| Category | Tools | Purpose |
|---|---|---|
| Workflow execution | get_workflow_tools() |
Run code reviews, security audits, refactoring, and other AI-assisted workflows |
| Documentation system | get_help_tools() |
Progressive help lookup, template maintenance, and project help initialization |
| Session utilities | get_utility_tools() |
Authentication status, telemetry stats, interaction level control, context management |
| Personal memory | get_personal_memory_tools() |
Store and recall decisions across sessions in ~/.attune/memory/ |
| Structured memory | get_memory_tools() |
Pattern storage and cross-agent coordination through the attune-ai memory module |
Protocol compliance
The server implements standard MCP resource types:
- Workflows resource (
attune://workflows) — JSON list of available automation workflows - Authentication config (
attune://auth/config) — Current auth strategy and subscription tier - Telemetry data (
attune://telemetry) — Cost tracking and performance metrics
Rate limiting protects against tool abuse with configurable call windows, defaulting to 60 calls per 60-second window.