Anvyon Agentic

Open-source multi-agent framework for task decomposition, delegation, and execution.

Anvyon Agentic is a Python library for building multi-agent systems that decompose, delegate, and execute complex tasks. It provides a simple yet powerful foundation for agentic workflows with pluggable LLM providers.

Features

• Zero required dependencies — core library uses only Python stdlib
• Task decomposition — break complex tasks into manageable subtasks via LLM
• Agent registry — register, list, and find agents by skill
• Basic delegation — create and manage delegation contracts
• Executor routing — route tasks to the right execution path (LLM or tool)
• Built-in tools — calculate, datetime, text_transform, json_parse, web_search
• Pluggable LLM providers — bring your own LLM or use the built-in Anthropic provider

Looking for advanced features? For advanced capabilities that enable creating more efficient, less cost workflows we have embedded operations and OB principles into this. Additional features like PERT analysis, Monte Carlo simulation, RL optimization, trust calibration, oversight/governance, delegation memory, and human-in-the-loop approval are available in the Anvyon Enterprise package. Contact anvyon1@gmail.com and visit www.anvyon.com for more details.

Installation

pip install anvyonagentic

With Anthropic (Claude) support:

pip install anvyonagentic[anthropic]

For development:

pip install anvyonagentic[dev]

Quick Start

from anvyonagentic import Task, AgentRegistry, BasicOrchestrator, ToolRegistry

tasks = [
    Task(id="T1", description="Parse user input",
         optimistic=0.5, most_likely=1.0, pessimistic=2.0),
    Task(id="T2", description="Process data",
         optimistic=1.0, most_likely=2.0, pessimistic=3.0,
         depends_on=["T1"]),
]

registry = AgentRegistry()
agents = registry.find_by_skill("reasoning")
print(f"Found agent: {agents[0].name}")

orchestrator = BasicOrchestrator()
results = orchestrator.run(tasks)
print(results)

tools = ToolRegistry()
result = tools.execute("calculate", {"expression": "2 + 2 * 3"})
print(result)

Architecture

Modules

Default Agents

Advanced Usage

Custom Agents

from anvyonagentic import AgentRegistry, WorkerAgent

registry = AgentRegistry()

registry.register(WorkerAgent(
    name="sql_expert",
    skills=["sql", "database", "query_optimization"],
    agent_type="tool",
    description="Specialized SQL and database agent"
))

agents = registry.find_by_skill("database")
print(f"Found agent: {agents[0].name}")

Built-in Tools

from anvyonagentic import ToolRegistry

tools = ToolRegistry()

result = tools.execute("calculate", {"expression": "2 + 2 * 3"})
print(result)  # {"success": True, "data": {"result": 8}}

result = tools.execute("datetime", {"format": "iso"})
result = tools.execute("text_transform", {"text": "hello", "operation": "upper"})

Enterprise Features

For advanced capabilities that enable creating more efficient, less cost workflows we have embedded operations and OB principles into this. Additional features like PERT analysis, Monte Carlo simulation, RL optimization, trust calibration, oversight/governance, delegation memory, and human-in-the-loop approval are available in the Anvyon Enterprise package.

Contact anvyon1@gmail.com and visit www.anvyon.com for more details.

Contributing

Contributions are welcome. Please open an issue first to discuss what you would like to change.

git clone https://github.com/anvyon1/Anvyonagentic.git
cd anvyonagentic
pip install -e ".[dev]"
pytest

License

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). See LICENSE for the full text.