# Dedalus Labs

> Production-grade SDK for building model-agnostic AI agents powered by MCP with secure, multi-tenant authentication.

Dedalus Labs provides a production-grade SDK for building model-agnostic AI agents powered by the Model Context Protocol (MCP). The platform enables developers to orchestrate agents with any LLM provider and connect to hosted MCP services instantly, all while maintaining enterprise-grade security with multi-tenant authentication. Backed by Y Combinator, Dedalus Labs offers a drop-in MCP gateway that handles hosting, scaling, and model hand-offs.

- **Universal Model Access** allows developers to instantly swap between models like OpenAI GPT, Anthropic Claude, Google Gemini, DeepSeek, and Mistral with a single line of code change.

- **Hosted MCP Marketplace** provides access to pre-built MCP servers for services like GitHub, Slack, Notion, Linear, and Brave Search that can be connected instantly without complex configuration.

- **Mix-and-Match Tooling** enables combining marketplace tools with custom local tools in the same agent workflow, supporting functions like file search, calendar, email, and database operations.

- **Zero Secret Leakage Security** ensures credentials never leave your local machine - remote MCP servers can authenticate using local credentials without ever seeing the secrets themselves.

- **Universal OAuth** provides one authentication layer for every MCP server, eliminating the need for custom OAuth flows per service.

- **One-Click Deployment** allows developers to deploy MCP servers to the cloud in seconds with global distribution across 300+ edge locations for Pro users.

- **Real-Time Monitoring Dashboard** tracks balance, request counts, and cost breakdowns per request with configurable log retention.

- **Marketplace Publishing** lets developers publish their own MCP servers and earn 80% of revenue when others use them.

- **Wingman TUI** is a model-agnostic, MCP-native terminal copilot for any use case with multi-model support, checkpoints, and diff previews.

To get started, install the Dedalus SDK for Python or TypeScript, initialize the AsyncDedalus client and DedalusRunner, then call runner.run() with your input, model preferences, MCP servers, and custom tools. The platform offers a free tier with 50 tool calls per month to begin building agents immediately.

## Features
- Universal model access across OpenAI, Anthropic, Google, DeepSeek, Mistral
- Hosted MCP marketplace with pre-built servers
- Mix-and-match local and marketplace tooling
- Zero secret leakage security architecture
- Universal OAuth for all MCP servers
- Client-side credential injection
- One-click cloud deployment
- Global edge distribution (300+ locations for Pro)
- Real-time usage monitoring dashboard
- Marketplace publishing with 80% revenue share
- Python and TypeScript SDK support
- Structured output support
- Universal streaming
- Wingman TUI copilot
- BYOK for model providers
- External MCP URLs support
- RBAC and team management (Enterprise)
- Audit logs (Enterprise)
- Custom SLA (Enterprise)

## Integrations
OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Mistral, GitHub, Slack, Notion, Linear, Brave Search, Twitter API

## Platforms
WEB, API, DEVELOPER_SDK

## Pricing
Open Source, Free tier available

## Links
- Website: https://www.dedaluslabs.ai
- Documentation: https://docs.dedaluslabs.ai/
- Repository: https://github.com/dedalus-labs/dedalus-sdk-python
- EveryDev.ai: https://www.everydev.ai/tools/dedalus-labs