Auth0 AI Agent Hackathon Submission
This guide walks you through testing the MCS Auth Bridge in three environments: a local CLI agent, an OpenWebUI tool, and a Claude Code Desktop (CoWork) skill. All three use the same SDK, the same Auth0 Token Vault integration, and the same LinkAuth device-flow UX.
Time estimate: ~30 min for Auth0 setup + ~10 min per use case.
| Feature | MCS Sandbox | MCP (Standard) | E2B | Vercel AI SDK |
|---|---|---|---|---|
| Tool Visibility | Dynamic (Lifecycle-aware) | Static (Always visible) | Static | Static |
| Backend Portability | Yes (Docker, SSH, etc.) | Server-specific | E2B Cloud Only | N/A |
| Environment State | Persistent (Volumes) | Ephemeral | Ephemeral | N/A |
| Loop Protection | Smart Fingerprinting | Up to client | Up to client | Fixed step limit |
| Truncation Handling | Auto-detected by Driver | Ignored / Client's job | Up to client | Up to client |
| Client Integration | ~3 lines of config | SDK + Transports | SDK | Framework specific |
| { | |
| "openapi": "3.0.3", | |
| "info": { | |
| "title": "Context7 Documentation API", | |
| "description": "Search programming libraries and retrieve up-to-date documentation and code examples. Use search_libraries first to resolve a library name to a Context7 library ID, then use query_documentation to fetch relevant docs.", | |
| "version": "2.0.0", | |
| "contact": { | |
| "name": "Context7", | |
| "url": "https://context7.com" | |
| } |
| """ | |
| Supabase RLS (Row Level Security) Audit Script | |
| ================================================ | |
| Testet ob die RLS-Policies einer Supabase-Instanz korrekt konfiguriert sind. | |
| Verwendet nur den öffentlich sichtbaren anon-Key (kein Service-Key nötig). | |
| """ | |
| import httpx | |
| import json | |
| import sys |
| import json | |
| import os | |
| import re | |
| from typing import Any, Optional | |
| import litellm | |
| from litellm import completion, completion_cost | |
| from litellm.caching.caching import Cache | |
| from litellm.main import ModelResponse, Usage | |
| from loguru import logger |
| #!/usr/bin/env python3 | |
| """ | |
| tau2_summarizer.py | |
| ------------------ | |
| Summarize Tau²-Bench result files (JSON) for Dual-Control, Retail runs. | |
| - Reads a Tau² result JSON (created via `--save-to ...`) | |
| - Aggregates per-task token usage and durations | |
| - Computes Pass@1, totals, and normalized "per interaction" metrics (Dual = 2 * tasks) | |
| - Optionally computes costs if prices are provided (USD per 1M tokens for input/output) |
| import json | |
| import os | |
| import re | |
| from typing import Any, Optional | |
| import litellm | |
| from litellm import completion, completion_cost | |
| from litellm.caching.caching import Cache | |
| from litellm.main import ModelResponse, Usage | |
| from loguru import logger |
| { | |
| "openapi": "3.0.0", | |
| "info": { | |
| "title": "Context7 API", | |
| "version": "1.0.0", | |
| "description": "LLM-optimized interface for accessing Context7 developer documentation." | |
| }, | |
| "servers": [ | |
| { | |
| "url": "https://context7.com/api/v1" |
| from appwrite.client import Client | |
| from appwrite.services.databases import Databases | |
| from appwrite.services.storage import Storage | |
| from appwrite.query import Query | |
| ... | |
| app_database = Databases(appwrite_client) | |
| result = database.list_documents( | |
| database_id=database_id, |
| from appwrite.client import Client | |
| from dotenv import load_dotenv | |
| def init_appwrite_client() -> Client: | |
| ''' | |
| Init Appwrite client and return Databases object | |
| ''' | |
| client = Client() | |
| client.set_endpoint(os.getenv("APPWRITE_ENDPOINT")) | |
| client.set_project(os.getenv("APPWRITE_PROJECT_ID")) |
