gridwork-aiact
# gridwork-aiact
MCP server that scans codebases for AI system usage and generates EU AI Act compliance reports.
**EU AI Act high-risk deadline: August 2, 2026.**
## What it does
- Scans project files for AI libraries, APIs, and frameworks
- Detects 30+ AI systems: OpenAI, Anthropic, Google, PyTorch, TensorFlow, HuggingFace, LangChain, and more
- Classifies risk level per EU AI Act (unacceptable / high / limited / minimal)
- Identifies compliance gaps against Annex III categories
- Generates formal inventory documents for compliance records
- Counts days until the August 2, 2026 deadline
## Install
```bash
npx gridwork-aiact
```
## MCP tools
| Tool | Description |
|------|-------------|
| `scan_project` | Full scan with risk classification and compliance gap analysis |
| `quick_check` | Fast count of AI systems without full analysis |
| `generate_inventory` | Formal EU AI Act inventory document |
| `classify_system` | Classify risk for a specific AI system by use case |
## Claude Desktop config
```json
{
"mcpServers": {
"gridwork-aiact": {
"command": "npx",
"args": ["-y", "gridwork-aiact"]
}
}
}
```
## License
MIT — [Gridwork](https://thegridwork.space)
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose: quick_check is a fast scan, scan_project is a full scan with report generation, generate_inventory produces a formal document, and classify_system classifies risk level. No overlap.
All tool names follow a consistent verb_noun pattern (quick_check, scan_project, generate_inventory, classify_system). The convention is uniform and predictable.
4 tools is well-scoped for the domain of EU AI Act compliance. Each tool serves a distinct and necessary function without redundancy or excessive granularity.
The tool set covers the core compliance workflow: fast scan, full scan, inventory generation, and risk classification. Minor gaps exist, such as the absence of tools for updating or editing generated inventories, but the essential operations are present.