SolaceLabs/solace-agent-mesh

⚠️ Solace Agent Mesh in Python is now deprecated

👋 Thank you to everyone who has built with Solace Agent Mesh! This Python version is now deprecated — it's no longer under active development and won't receive new features, bug fixes or security updates.

🚀 Check out the new version of Solace Agent Mesh → https://docs.solace.com/Agent-Mesh/agent-mesh.htm

🖥️ A free edition of the Solace Agent Mesh desktop app is available: https://solace.com/products/agent-mesh/download

This repository will be archived (read-only) but stays available for reference, so your existing links and installs keep working.

Key Features • Quickstart • Next Steps • Docs

Solace Agent Mesh is a framework that supports building AI applications where multiple specialized AI agents work together to solve complex problems. It uses the event messaging of Solace Platform for true scalability and reliability.

With Solace Agent Mesh (SAM), you can create teams of AI agents, each having distinct skills and access to specific tools. For example, you could have a Database Agent that can make SQL queries to fetch data or a MultiModal Agent that can help create images, audio files and reports.

The framework handles the communication between agents automatically, so you can focus on building great AI experiences.

SAM creates a standardized communication layer where AI agents can:

• Delegate tasks to peer agents
• Share data and artifacts
• Connect with diverse user interfaces and external systems
• Execute multi-step workflows with minimal coupling

SAM is built on top of the Solace AI Connector (SAC) which allows Solace Platform Event Brokers to connect to AI models and services and Google's Agent Development Kit (ADK) for AI logic and tool integrations.

The result? A fully asynchronous, event-driven and decoupled AI agent architecture ready for production deployment. It is robust, reliable and easy to maintain.

• Multi-Agent Event-Driven Architecture – Agents communicate via the Solace Event Mesh for true scalability
• Agent Orchestration – Complex tasks are automatically broken down and delegated by the Orchestrator agent
• Flexible Interfaces – Integrate with REST API, web UI, Slack, or build your own integration
• Extensible – Add your own agents, gateways, or services with minimal code
• Agent-to-Agent Communication – Agents can discover and delegate tasks to each other seamlessly using the Agent2Agent (A2A) Protocol
• Dynamic Embeds – Embed dynamic content like real-time data, calculations and file contents in responses

📚 Want to know more? Check out the full Solace Agent Mesh documentation.

Set up Solace Agent Mesh in just a few steps.

To run Solace Agent Mesh locally, you'll need:

• Python 3.10.16 – 3.13.x
• pip (comes with Python)
• OS: MacOS, Linux, or Windows (with WSL)
• LLM API key (any major provider or custom endpoint)

To quickly setup and customize your Agent Mesh, check out the Vibe Coding Quickstart Guide. This guide walks you through the essential steps to get Solace Agent Mesh up and running with minimal effort.

mkdir my-sam && cd my-sam

python3 -m venv .venv && source .venv/bin/activate

Check if you have a version of SAM already installed.

sam -v

If you have an earlier version, uninstall it and start from scratch:

pip3 uninstall solace-agent-mesh

Note: Optionally, you can try to upgrade versions but this action is not officially supported at this time. (pip3 install --upgrade solace-agent-mesh)

If no previous version exists, install the latest version with:

pip3 install solace-agent-mesh

sam init --gui

Note: This initialization UI runs on port 5002

sam run

Open the Web UI at http://localhost:8000 for the chat interface and ask a question

sam add agent --gui

sam plugin add <your-component-name> --plugin <plugin-name>

Solace Agent Mesh provides a "Universal A2A Agent Host," a flexible and configurable runtime environment built by integrating Google's Agent Development Kit (ADK) with the Solace AI Connector (SAC) framework.

The system allows you to:

• Host AI agents developed with Google ADK within the SAC framework
• Define agent capabilities (LLM model, instructions, tools) primarily through SAC YAML configuration
• Use Solace Platform as the transport for standard Agent-to-Agent (A2A) protocol communication
• Enable dynamic discovery of peer agents running within the same ecosystem
• Allow agents to delegate tasks to discovered peers via the A2A protocol over Solace
• Manage file artifacts using built-in tools with automatic metadata injection
• Perform data analysis using built-in SQL, JQ, and visualization tools
• Use dynamic embeds for context-dependent information resolution

• SAC handles broker connections, configuration loading, and component lifecycle
• ADK provides the agent runtime, LLM interaction, tool execution, and state management
• A2A Protocol enables communication between clients and agents, and between peer agents
• Dynamic Embeds allow placeholders in responses that are resolved with context-dependent information
• File Management provides built-in tools for artifact creation, listing, loading, and metadata handling

Want to go further? Here are some hands-on tutorials to help you get started:

Solace Agent Mesh is built with the help of our amazing community. Thanks to everyone who has contributed ideas, code and time to make this project better!

View the full list of contributors here: GitHub Contributors 💚

Looking to contribute? Check out CONTRIBUTING.md to get started and see how you can help!

This project is licensed under the Apache 2.0 License. See the full license text in the LICENSE file.

This project uses pytest for testing. You can run tests using either hatch or pytest directly.

The recommended way to run tests is through the hatch environment, which ensures all dependencies are managed correctly.

# Run all tests
hatch test

# Run tests with tags
hatch test -m "<tag>"

If you prefer to use pytest directly, you must first install the project with its test dependencies.

# Install the project in editable mode with the 'test' extras
pip install -e .[test]

# Run all tests
pytest