Microsoft Agent Framework + Neo4j Integration

Microsoft Agent Framework is Microsoft’s open-source SDK for building production AI agents in Python and .NET — single agents, multi-agent workflows, and hosted-agent deployment. Neo4j is the graph database and knowledge layer that grounds those agents in connected enterprise data — relationships, hierarchies, and multi-hop paths the model can actually reason over.

Better together: Agent Framework gives you agents that collaborate. Neo4j gives those agents shaped, connected context instead of flattened retrieval.

Agent Framework is cross-vendor — Azure OpenAI, OpenAI, Anthropic, local models, and more. Both examples below use the Microsoft Foundry deployment from ../microsoft-foundry/ — the local example calls the project endpoint directly, the hosted example registers a containerised agent against the same project. One deployment, one .env, both use cases.

Why multi-agent + graph

Investment research is multi-step: discover, profile, traverse the network, read the news, synthesize. A single fat agent with twenty tools loses focus — splitting it into a Database agent and an Analyst produces sharper results, and Agent Framework makes that composition trivial: coordinator.tools = [db.as_tool(), analyst.as_tool()].

Architecture

flowchart LR
    user["User"] --> coord["Coordinator agent<br/>(FoundryChatClient)"]
    coord -->|as_tool| db["Database agent<br/>(10 Neo4j function tools)"]
    coord -->|as_tool| analyst["Analyst agent<br/>(synthesis only)"]
    db -->|neo4j Bolt driver| neo4j[("Neo4j Aura<br/>or self-managed")]
    coord -.model.-> foundry["Foundry project<br/>(../microsoft-foundry/)"]

Both examples implement the same multi-agent graph. The local example runs your code; the hosted example runs that code inside Foundry’s managed runtime.

Quick start

az login
cd microsoft-foundry/infra && ./deploy.sh    # one-time, opt in to Foundry
cd ../../microsoft-agent-framework/examples/multi-agent && uv run multi_agent_neo4j.py

Defaults connect to the public companies Neo4j demo graph and use the Foundry project provisioned by deploy.sh.

Examples

Example What it shows Folder

Multi-agent (local)

Coordinator + Database + Analyst via as_tool(), function tools to Neo4j

examples/multi-agent

Foundry-hosted multi-agent

Same multi-agent graph packaged as a Foundry hosted agent

examples/foundry-hosted

Why host on Foundry?

Hosted agents in Foundry Agent Service let you ship the same Agent Framework code to a managed runtime. From the official hosted-agents concepts page:

  • Bring your own code — Agent Framework, LangGraph, Semantic Kernel, or custom; the platform doesn’t care.

  • Dedicated agent identity — a Microsoft Entra ID is auto-created at deploy and used by your agent at runtime to call models, tools, and downstream Azure services.

  • Per-session VM-isolated sandboxes — $HOME and /files persist across turns and idle (15-min idle timeout, 30-day session lifetime).

  • Versioning — immutable agent versions with weighted traffic split for canary and blue-green rollouts.

  • Scale-to-zero — the platform handles container lifecycle, scaling, and observability via Application Insights.

  • Foundry portal integration — playground, version management, and traces, with no extra wiring.

References

Microsoft Agent Framework

Neo4j