Open, modular, and language-agnostic data and agent interoperability for "truly autonomous" AI workflows
Agent Data Shuttle (ADS) is an open-source protocol for connecting AI agents to realtime data sources - across languages, runtimes, and platforms.
ADS enables seamless autonomous invocation of your AI agents, allowing them to react to events, monitor systems, and interact with tools in real-time without human intervention.
- 📚 Read the Documentation for concepts, architecture, and usage guides
- 🚀 Try the Quickstart Projects
- 💻 Build with our SDKs:
- 🧩 Integrate with n8n and other workflow tools
| Repository | Description |
|---|---|
ads-bridge |
Core bridge server to communicate between ADS Publisher and Subscribers |
ads-documentation |
Docs, concepts, and architecture |
ads-example-projects |
Example projects and quickstarts |
n8n-nodes |
n8n integration nodes for ADS |
python-sdk |
Python SDK for ADS |
typescript-sdk |
TypeScript SDK for ADS |
We welcome issues, pull requests, and design discussions. If you’d like to add support for another language, tool, or framework, open a discussion first so we can align on the design!
Agent Data Shuttle is released under the Apache-2.0 license and maintained by a growing community of AI and data workflow enthusiasts. If your organization needs robust, direct, and secure agent-to-tool communication, or you want to avoid writing wrappers, we’d love to have you involved!
MCP: Model Context Protocol (MCP) uses a “pull” approach - data is fetched when explicitly requested. This is perfect for conversational scenarios and on-demand information retrieval.
ADS: Agent Data Shuttle (ADS) uses a “push” approach - data flows automatically when events occur. This enables real-time updates, reactive agent invocation, and truly autonomous AI behavior, empowering agents to act without human prompting.
| MCP (Pull) | ADS (Push) | |
|---|---|---|
| Data Flow | Pull-based (Manual) | Push-based (Automatic) |
| Trigger | User/Agent Query | Event Occurrence |
| Use Case | On-demand agentic flows | Real-time agentic flows |
| Latency | Query time | Near real-time |
Choose MCP when:
- You need on-demand agent invocation
- Your AI agent needs to answer specific questions
- Query-response is your primary interaction model
Choose ADS when:
- You need real-time agent invocations based on external events
- Your AI agent needs to stay current with system events
- Proactive monitoring and response is crucial - true “autopilot” mode
💡 MCP × ADS is better than either alone.
- Use MCP for your agent’s active inquiries and deep dives.
- Use ADS for keeping your agent autonomously aware and responsive.
Combine both for an AI that can both investigate (MCP) and independently react (ADS).