Clawdette

Clawdette

Your AI-Powered Multi-Channel Agent

Intelligent conversations powered by Claude AI with 27 built-in tools, MCP integration, and enterprise-grade security.

Features

AI Conversations

Powered by Claude (Anthropic) - intelligent, context-aware conversations with multi-turn support.

27 Built-in Tools

DateTime, Web Search, TTS, Bash, Email, Calendar, Image Generation, Scheduling and more - ready to use.

MCP Integration

Extend with any MCP server - connect external APIs and services seamlessly.

Enterprise Security

Sandbox VM isolation, external API key management, and per-tenant workspace separation.

Multi-Tenant

Isolated workspaces with individual settings, tools and permissions per user.

Calendar Integration

Connect Google, Microsoft, or Apple calendars - manage events, check availability, and schedule directly.

Tools & Capabilities

27 native tools across 6 categories, plus unlimited MCP extensions.

Standard Tools

4 tools

DateTime, Web Fetch, Memory Search, Memory Get

API-Powered

5 tools

Web Search (Brave), TTS (OpenAI), Image Analysis, Image Generate, Vision

Sandbox

6 tools

Bash, Read, Write, Edit, File Share, Browser

Email

3 tools

Fetch, Send, Search

Calendar

6 tools

List, Search, Create, Update, Delete, Availability

Scheduling

3 tools

Schedule Job, List Schedules, Delete Schedule

Security by Design

AI agents with unchecked access are a liability. Clawdette isolates every layer by default.

Sandbox VM Isolation

All code execution runs in an isolated virtual machine, never on the application server.

External API Key Management

API keys are stored server-side and injected at execution time. The AI model never sees credentials.

Per-Tenant Workspace Isolation

Each user operates in a fully isolated workspace with separate settings, tools, and permissions.

Admin-Controlled Tool Access

Administrators enable or disable tools per tenant. No tool runs without explicit authorization.

No Credential Exposure

API keys stored securely, never passed in prompts or tool descriptions. Zero prompt injection risk for credentials.

Why This Matters

Recent CVEs in open-source AI agents have exposed thousands of instances to credential theft and host compromise. Common architectural flaws include:

Credentials embedded in system prompts, extractable via prompt injection
No sandbox isolation - code execution directly on the host system
Malicious plugins with unrestricted filesystem and network access
No per-user isolation - shared state across all users