Hyland recently announced the Enterprise Agent Mesh as part of the Enterprise Context Engine. This initiative shows how Large Language Models (LLMs) can delegate work to specialized agents in a coordinated way.
To help Alfresco Community developers explore this concept, we’ve built a simple proof-of-concept project: simple-alfresco-agent-mesh
The goal of this project is to demonstrate how an Agent Mesh can orchestrate multiple Model Context Protocol (MCP) servers to perform different tasks against an Alfresco repository.
Background: From Single MCP to Mesh
If you followed our earlier Alfresco MCP Lab, you know how a single MCP server can expose Alfresco operations to an LLM client.
The new Agent Mesh takes this one step further: instead of a single MCP, we now have a router that decides which MCP to call depending on the request. For example:
- A Docs MCP handles searches and content queries
- An Audit MCP handles audit trail queries
This aligns perfectly with Hyland’s Agent Mesh vision: multiple agents specialized in their domain, orchestrated dynamically.
Project Overview
The repository includes a router MCP server built with LangGraph. The router analyzes the user request and forwards it to the appropriate Alfresco MCP server.
The mesh currently integrates two MCP servers provided by the community:
- alfresco-mcp by @jottley (Java/Spring Boot)
- python-alfresco-mcp-server by @stevereiner (Python)
Both connect directly to Alfresco Community and expose different sets of tools.
Note: This version is based on LangGraph for routing, but a LlamaIndex-based version is also in progress: stay tuned!
Get Started
Clone the repo and spin up the stack with Docker Compose:
git clone https://github.com/aborroy/simple-alfresco-agent-mesh.git
cd simple-alfresco-agent-mesh
docker compose up --buildOnce running, you can connect with the MCP client and try the same queries.
docker compose -f mcp-client-cmd/compose.yaml run --rm mcp-clientSearching documents in Alfresco
> I need to search for documents in Alfresco. Can you search for:
- Documents containing "budget"
- Maximum 5 results
The router selects the Docs MC* and returns a list of matching documents (e.g., budget.xls, Meeting Notes, etc.).
Listing audit applications
> List Alfresco audit apps
This time, the router forwards the request to the Audit MCP, which returns configured audit apps (tagging, search).
Retrieving audit entries
> List Alfresco audit entries for audit app "search"
Again, the request is routed to the Audit MCP, returning the latest audit entries for the search app.
These examples show the MCP selection in action: the mesh automatically chooses the right agent for the task without requiring manual intervention.
Why This Matters
This project demonstrates how AI-powered orchestration can bring together different Alfresco services under a single conversational interface. Developers can:
- Extend the mesh with new MCP servers (e.g., metadata extraction, workflows)
- Experiment with hybrid orchestration (LangGraph today, LlamaIndex tomorrow)
- Contribute tools and patterns for the Alfresco developer community
Credits
This project builds on the great work of the community. Thanks to their contributions, we can now show how an Agent Mesh can be applied to Alfresco Community.
Next step: we’ll publish the LlamaIndex version of the router soon, so developers can compare orchestration approaches and extend them further.
