marc-shade/coral-tpu-mcp
If you are the rightful owner of coral-tpu-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.
Coral TPU MCP Server provides fast local machine learning inference for agentic systems using Google Coral Edge TPU.
Coral TPU MCP Server
Google Coral TPU integration for edge ML inference.
Part of the Agentic System - a 24/7 autonomous AI framework with persistent memory.
Fast local ML inference for agentic systems using Google Coral Edge TPU.
Features
- Image Classification: 1000 ImageNet classes with MobileNet/EfficientNet
- Visual Embeddings: Generate embeddings for image similarity
- Text Embeddings: CPU-based text embeddings with small models
- Importance Scoring: Score memory importance for prioritization
- Anomaly Detection: Detect anomalies in data patterns
- Face Detection: Detect faces in images
MCP Tools
| Tool | Description |
|---|---|
classify_image | Classify image using MobileNet V2 |
get_visual_embedding | Generate visual embedding for an image |
get_text_embedding | Generate text embedding (CPU) |
score_importance | Score memory importance (0-1) |
detect_anomaly | Detect anomalies in data |
detect_faces | Detect faces in an image |
Models
mobilenet_v2- Image classification (224x224 input)efficientnet_s- Visual embeddings and classificationface_detection- Face detection model
Requirements
- Python 3.10+
- Google Coral Edge TPU USB Accelerator
- pycoral and tflite-runtime
- mcp SDK
Installation
pip install -e .
Usage
python -m coral_tpu_mcp.server
Hardware
Requires a Google Coral Edge TPU (USB or PCIe) for accelerated inference. Falls back to CPU for text embeddings.
Integration
Provides fast local inference for:
- Memory importance scoring
- Visual episode encoding
- Image-based context understanding
License
MIT
Part of the MCP Ecosystem
This server integrates with other MCP servers for comprehensive AGI capabilities:
| Server | Purpose |
|---|---|
| enhanced-memory-mcp | 4-tier persistent memory with semantic search |
| agent-runtime-mcp | Persistent task queues and goal decomposition |
| agi-mcp | Full AGI orchestration with 21 tools |
| cluster-execution-mcp | Distributed task routing across nodes |
| node-chat-mcp | Inter-node AI communication |
| ember-mcp | Production-only policy enforcement |
See agentic-system-oss for the complete framework.