nerve is a Python-based Agent Development Kit (ADK) for building, running, evaluating, and orchestrating LLM-powered agents. It targets technical users seeking programmable, auditable, and reproducible automation, offering a declarative YAML-based approach for agent definition and workflow composition.

How It Works

Nerve utilizes a declarative YAML format to define agents, specifying system prompts, tasks, tools, and variables. It integrates with LiteLLM for LLM-agnosticism, supporting numerous providers. Agents can leverage built-in tools (shell commands, Python functions) and custom remote tools. A key feature is native Message Communication Protocol (MCP) support, enabling agents to act as both clients and servers for complex orchestration and team-like interactions.

Quick Start & Requirements

• Install with: pip install nerve-adk
• Create an agent with: nerve create new-agent
• Run an agent with: nerve run new-agent
• Documentation: https://github.com/evilsocket/nerve/blob/main/docs/index.md
• Examples: https://github.com/evilsocket/nerve/tree/main/examples

Highlighted Details

• Declarative agent definition via YAML.
• LLM-agnostic via LiteLLM support.
• Native MCP client and server capabilities for agent orchestration.
• Evaluation mode for benchmarking agents with test cases.
• Workflow composition for multi-step automations.

Maintenance & Community

• Active development with CI pipelines.
• Community support via Discord.
• Latest release available on PyPI and Docker Hub.

Licensing & Compatibility

• License: GPL-3.0
• GPL-3.0 is a strong copyleft license, requiring derivative works to also be open-sourced under GPL-3.0. This may restrict commercial use or integration into closed-source proprietary software without careful consideration.

Limitations & Caveats

The GPL-3.0 license imposes significant restrictions on commercial use and integration with proprietary software due to its strong copyleft nature.