Agent Haymaker

Universal workload orchestration platform for Azure and M365 telemetry generation.

Agent Haymaker deploys and manages workloads – specialized agents that generate realistic telemetry in Azure tenants and M365 environments. It provides a single CLI and a standard interface so that any workload, whether it simulates Azure infrastructure scenarios or M365 knowledge workers, plugs in and behaves the same way.

Guides

LLM Provider Configuration – Set up Anthropic Claude, Azure OpenAI, or Azure AI Foundry for AI-powered workloads
Creating Workloads – Build a custom workload package that implements the WorkloadBase interface
CLI Reference – Complete reference for every haymaker command and option

Quick Start

Install

pip install agent-haymaker

Install a workload

# From a git repository
haymaker workload install https://github.com/rysweet/haymaker-m365-workloads

# From a local path
haymaker workload install ./my-workload

Deploy

# Deploy M365 knowledge workers
haymaker deploy m365-knowledge-worker --config workers=25 --config department=sales

# Deploy an Azure infrastructure scenario
haymaker deploy azure-infrastructure --config scenario=linux-vm-web-server

Manage

haymaker status dep-abc123          # Check status
haymaker logs dep-abc123 --follow   # Stream logs
haymaker stop dep-abc123            # Stop a deployment
haymaker start dep-abc123           # Resume a stopped deployment
haymaker cleanup dep-abc123         # Delete all resources
haymaker list --status running      # List active deployments

See the CLI Reference for the full set of commands and options.

Architecture

+------------------------------------------------------------------+
|                    AGENT HAYMAKER PLATFORM                        |
|                                                                   |
|  WorkloadBase (interface)         Universal CLI                   |
|  +-- deploy()                     +-- haymaker deploy             |
|  +-- get_status()                 +-- haymaker status             |
|  +-- stop()                       +-- haymaker stop               |
|  +-- start()                      +-- haymaker start              |
|  +-- cleanup()                    +-- haymaker cleanup            |
|  +-- get_logs()                   +-- haymaker logs               |
|                                                                   |
|  Workload Registry    State Storage    Credential Management      |
+------------------------------------------------------------------+
                              |
                   implements | WorkloadBase
                              v
+------------------------------------------------------------------+
|                    WORKLOAD PACKAGES                              |
|                                                                   |
|  haymaker-azure-workloads       haymaker-m365-workloads           |
|  +-- Azure infrastructure       +-- Knowledge workers             |
|  +-- Goal-seeking agents        +-- M365 operations               |
|  +-- Scenario execution         +-- Entra identity mgmt           |
+------------------------------------------------------------------+

The platform provides the WorkloadBase interface, a registry for discovering installed workloads, state storage for tracking deployments, and credential management via Azure Key Vault. Workload packages are separate repositories that implement the interface and can be installed via haymaker workload install.

LLM Integration

Workloads can use the built-in multi-provider LLM abstraction layer for AI-powered content generation, adaptive agent behavior, and intelligent operations. See the LLM Provider Configuration guide for details.

from agent_haymaker.llm import create_llm_client, LLMConfig, LLMMessage

config = LLMConfig.from_env()
client = create_llm_client(config)

messages = [LLMMessage(role="user", content="Write a professional email about Q4 results")]
response = client.create_message(messages, system="You are a business writer.")

Supported providers: Anthropic Claude, Azure OpenAI, and Azure AI Foundry.

Workload Repositories

Workload	Description	Repository
`azure-infrastructure`	Azure scenario execution with goal-seeking agents	haymaker-azure-workloads
`m365-knowledge-worker`	M365 knowledge worker simulation	haymaker-m365-workloads

Development

git clone https://github.com/rysweet/agent-haymaker
cd agent-haymaker
pip install -e ".[dev]"
pytest

License

MIT