XDS110 MCP Server
Enables Large Language Models to act as co-debuggers for Texas Instruments embedded systems by providing real-time variable monitoring, memory manipulation, and motor control analysis through OpenOCD proxy architecture. Supports debugging TI microcontrollers with XDS110 debug probes while working alongside Code Composer Studio.
README
XDS110 MCP Server
LLM Co-Debugger for TI Embedded Systems
An MCP (Model Context Protocol) server that enables Large Language Models to act as co-debuggers for Texas Instruments embedded systems. Provides real-time variable monitoring, memory manipulation, and analysis through OpenOCD proxy architecture.
Functionality
This MCP server enables LLMs to:
- Monitor 50+ motor control variables simultaneously with change detection
- Analyze root causes of motor humming, overcurrent faults, and calibration failures
- Write to debug structures and motor control parameters via direct memory access
- Work alongside Code Composer Studio through session handoff
- Provide domain expertise in motor control, FOC principles, and TI peripherals
Verified Hardware Support
- TI XDS110 Debug Probe (USB connection)
- TMS320F280039C microcontroller (F28xx series)
- Custom boards with XDS110 debugger integration
- PMSM motor control via DRV8323RH driver
- Direct memory access to union structures
Architecture Overview
graph TB
A[LLM Client<br/>Claude/GPT] <--> B[XDS110 MCP Server<br/>Python]
B <--> C[OpenOCD Proxy<br/>Multi-Client GDB]
C <--> D[XDS110 Debug Probe<br/>USB Hardware]
D <--> E[TMS320F280039C<br/>Target MCU]
B <--> F[Domain Knowledge<br/>Motor Control/FOC]
G[Code Composer Studio] -.-> C
subgraph "MCP Tools"
H[read_variables]
I[monitor_variables]
J[write_memory]
K[set_breakpoint]
L[analyze_motor_state]
end
B --> H
B --> I
B --> J
B --> K
B --> L
OpenOCD Multi-Client Proxy
Traditional debugging allows only one tool to connect to hardware. This solution uses OpenOCD's multi-client capabilities:
- OpenOCD connects to XDS110 using native driver support
- Multiple GDB clients can connect to OpenOCD simultaneously
- MCP server acts as intelligent GDB client for LLM interaction
- Code Composer Studio can still connect when needed (via session handoff)
Quick Start
Prerequisites
- Python 3.8+
- OpenOCD with XDS110 driver support
- XDS110 debug probe with firmware 2.3.0.11+
- TI target hardware (F280039C verified)
Installation
git clone https://github.com/yourusername/XDS110_MCP_server.git
cd XDS110_MCP_server
pip install -r requirements.txt
Hardware Verification
# Check XDS110 detection
lsusb | grep "0451:bef3"
# Should show: Bus XXX Device XXX: ID 0451:bef3 Texas Instruments, Inc.
# Test OpenOCD connection
openocd -f configs/xds110_f28039.cfg
Start MCP Server
# Start the MCP server
python -m xds110_mcp_server --config configs/f28039_config.json
# In another terminal, test with MCP Inspector
npx @modelcontextprotocol/inspector python -m xds110_mcp_server
MCP Tools Available to LLMs
| Tool | Description | Parameters | Example Use |
|---|---|---|---|
read_variables |
Read current values of motor control variables | variable_list, format |
Monitor motor state, position, currents |
monitor_variables |
Continuous monitoring with change detection | variables, duration, threshold |
Detect anomalies during motor operation |
write_memory |
Direct memory writes to structures | address, value, size |
Modify debug_bypass parameters |
set_breakpoint |
Conditional breakpoints with custom logic | address, condition, action |
Stop when motor state changes |
analyze_motor_state |
AI-powered motor control analysis | focus_area |
Diagnose humming, faults, tuning issues |
calibrate_motor |
Trigger calibration sequences | calibration_type |
Run commands 64-67 for motor setup |
Example LLM Interactions
Human: "My PMSM motor hums instead of spinning during bypass alignment"
LLM: I'll help diagnose this motor humming issue. Let me start by monitoring
the key variables during bypass alignment.
→ monitor_variables(['motorVars_M1.motorState', 'debug_bypass.bypass_alignment_called', 'motorVars_M1.Idq_out_A.value[0]'], duration=30)
I can see the bypass alignment is being called, but the D-axis current
(Idq_out_A[0]) remains at 0. This suggests missing current control
initialization. Let me check the debug_bypass structure...
→ read_variables(['debug_bypass.debug_enabled', 'debug_bypass.command.cmd'])
The debug_bypass.debug_enabled is 0. Let me enable it and set up proper
current control initialization:
→ write_memory(0x0000d3c0, 1, 1) # Enable debug bypass
→ write_memory(0x0000d3c5, 100, 2) # Set 0.1A current limit
Now let's monitor if the motor behavior improves...
Project Structure
XDS110_MCP_server/
├── README.md # This file
├── PRD.md # Product Requirements Document
├── LICENSE # MIT License
├── requirements.txt # Python dependencies
├── pyproject.toml # Python project configuration
│
├── xds110_mcp_server/ # Main MCP server package
│ ├── __init__.py # Package initialization
│ ├── server.py # Main MCP server implementation
│ ├── tools/ # MCP tool implementations
│ │ ├── __init__.py
│ │ ├── variable_monitor.py # Variable reading/monitoring tools
│ │ ├── memory_tools.py # Memory read/write tools
│ │ ├── breakpoint_tools.py # Breakpoint management tools
│ │ └── analysis_tools.py # Motor analysis and diagnostics
│ ├── gdb_interface/ # GDB protocol communication
│ │ ├── __init__.py
│ │ ├── gdb_client.py # GDB protocol implementation
│ │ └── openocd_manager.py # OpenOCD process management
│ ├── knowledge/ # Domain knowledge database
│ │ ├── __init__.py
│ │ ├── motor_control.py # Motor control expertise
│ │ ├── ti_peripherals.py # TI peripheral knowledge
│ │ └── fault_patterns.py # Common failure pattern recognition
│ └── utils/ # Utility functions
│ ├── __init__.py
│ ├── config.py # Configuration management
│ ├── logging.py # Logging setup
│ └── hardware_detect.py # Hardware detection utilities
│
├── configs/ # Configuration files
│ ├── xds110_f28039.cfg # OpenOCD configuration for F280039C
│ ├── f28039_config.json # Default server configuration
│ └── motor_variables.json # Motor control variable definitions
│
├── legacy_ti_debugger/ # Copied from working implementation
│ ├── framework/
│ │ └── ti_dss_adapter.py # Proven TI DSS connection logic
│ ├── working_memory_motor_control.py # Working motor control script
│ ├── motor_control.py # Clean entry point
│ └── js_scripts/ # JavaScript DSS debugging scripts
│ ├── connect_target_v2.js
│ ├── read_motor_vars_v1.js
│ └── monitor_alignment.js
│
├── tests/ # Test suite
│ ├── __init__.py
│ ├── test_mcp_server.py # MCP server tests
│ ├── test_gdb_interface.py # GDB interface tests
│ ├── test_hardware_integration.py # Hardware integration tests
│ └── fixtures/ # Test fixtures and mock data
│
├── docs/ # Documentation
│ ├── installation.md # Installation guide
│ ├── configuration.md # Configuration documentation
│ ├── api_reference.md # MCP API reference
│ ├── troubleshooting.md # Common issues and solutions
│ └── examples/ # Usage examples
│ ├── basic_debugging.md
│ ├── motor_tuning.md
│ └── advanced_analysis.md
│
└── scripts/ # Utility scripts
├── setup_openocd.sh # OpenOCD installation helper
├── test_hardware.py # Hardware connection test
└── validate_installation.py # Installation validation
Development Status
Completed (From Legacy ti_debugger)
- [x] Hardware Connection: XDS110 + F280039C proven working
- [x] Memory Access: Direct read/write to debug_bypass structure (0x0000d3c0)
- [x] Variable Reading: Comprehensive motor control variable access
- [x] Motor Control: PMSM motor control with DRV8323RH driver
- [x] Calibration: Automated calibration sequences (commands 64-67)
- [x] Domain Knowledge: Motor control, FOC principles, fault patterns
In Progress
- [ ] MCP Server Implementation: Converting ti_debugger logic to MCP framework
- [ ] OpenOCD Integration: Multi-client GDB proxy setup
- [ ] Tool Development: MCP tools for variable monitoring, memory access
- [ ] LLM Integration: Domain knowledge integration for intelligent analysis
Planned
- [ ] Session Management: CCS handoff and conflict resolution
- [ ] Advanced Analysis: Pattern recognition and fault diagnosis
- [ ] Performance Optimization: Sub-100ms variable read latency
- [ ] Documentation: Comprehensive guides and API reference
- [ ] Testing: Hardware integration and reliability tests
Use Cases
1. Motor Humming Diagnosis
Problem: PMSM motor hums during bypass alignment instead of spinning smoothly. Solution: LLM analyzes bypass alignment variables, identifies missing current control initialization, and suggests memory writes to fix the issue.
2. Real-Time Debugging Assistant
Problem: Complex motor control sequences with dozens of variables to monitor. Solution: LLM continuously monitors variables, detects anomalies, and provides contextual alerts with domain expertise.
3. Interactive Parameter Tuning
Problem: PID controller tuning requires iterative testing and analysis. Solution: LLM suggests parameter changes based on motor behavior description, applies changes via memory writes, and monitors results.
4. Automated Fault Analysis
Problem: Intermittent overcurrent faults that are difficult to debug manually. Solution: LLM sets intelligent breakpoints, monitors fault conditions, and analyzes patterns to identify root causes.
Technical Specifications
Performance Requirements
- Variable Read Latency: < 100ms per variable
- Monitoring Frequency: Up to 10Hz for critical variables
- Memory Footprint: < 50MB RAM usage
- Startup Time: < 5 seconds to ready state
- Connection Recovery: Auto-reconnect within 1 second
Hardware Requirements
- Debug Probe: XDS110 with firmware 2.3.0.11+
- Target MCU: TMS320F280039C (F28xx series support planned)
- Connection: USB 2.0+ for XDS110 probe
- Host OS: Linux (tested), Windows/macOS (planned)
Software Requirements
- Python: 3.8+ with asyncio support
- OpenOCD: Latest version with XDS110 driver
- MCP SDK: Python Model Context Protocol SDK
- Optional: Code Composer Studio for traditional debugging
Contributing
We welcome contributions to this embedded systems debugging MCP server:
High-Priority Areas
- Additional TI MCU support (F28xx series, C2000 family)
- Other debug probe support (J-Link, ST-Link)
- Windows/macOS compatibility
- Advanced motor control algorithms
- Performance optimizations
Getting Started
- Fork the repository
- Set up development environment:
pip install -r requirements-dev.txt - Run tests:
pytest tests/ - Submit pull request with comprehensive testing
Development Guidelines
- Follow PEP 8 style guidelines
- Add tests for new functionality
- Update documentation for user-facing changes
- Verify hardware compatibility before submitting
Resources & References
MCP Protocol
OpenOCD & Hardware
TI Documentation
License
MIT License - see LICENSE file for details.
Acknowledgments
- Texas Instruments: For XDS110 debug probe and comprehensive documentation
- OpenOCD Project: For multi-client debugging architecture
- Model Context Protocol Team: For the MCP framework enabling LLM tool integration
- Embedded Community: For sharing knowledge and debugging techniques
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.