Easy Agents: Automate Operations with Natural Language and MCP

Easy Agents swaps traditional scripts for natural language instructions and Model Context Protocol (MCP) tool calls. By leveraging large language models like Claude Code, it translates high-level objectives into executable workflows. The system is designed to handle complex tasks and multi-source data intelligently, allowing you to automate security investigations, code remediation, supply chain management, and team reporting. Standardized protocols streamline the integration of external tools. Because the automation logic is written in plain English, your team can build, understand, and scale workflows without the burden of a steep learning curve.

Simply define your agent's behavior in everyday language, and Easy Agents manages the messy, unstructured details of the workflow.

For example:

You are a security agent investigating alerts.

Main steps:

1. Locate the alert by its ID.
2. Perform a deep analysis of the event.
3. Generate a detailed report of your findings.
4. If the alert is verified as valid, post the report to the #security-alerts Slack channel.

Real-world use cases

Easy Agents integrates with various third-party MCP servers, including Panther security monitoring, VirusTotal, GitHub, Linear, and Slack.

Security alert investigation agent – This agent classifies incoming alerts by correlating events across different sources to assess risk levels and route them to the correct teams. It analyzes alert context, investigates related activities, and determines the appropriate severity and response.

App security agent – This agent monitors GitHub Advanced Security code scanning alerts and automatically generates pull requests to fix verified issues. It runs tests and linters to ensure code quality before assigning the relevant reviewers to the PR.

Supply chain security agent – This agent triages GitHub Dependabot alerts by evaluating the impact of vulnerabilities. It updates packages to secure versions, confirms that tests pass, and manages the entire pull request lifecycle.

Weekly summary agent – Every Sunday, this agent reviews the team’s projects and tickets in Linear. It compiles a report detailing active projects, work completed during the week, and individual contributions.

Claude Code provides the logic required for sophisticated task orchestration. It breaks down high-level goals into actionable plans, pivots when it encounters roadblocks, and applies critical thinking to complex problems. When paired with the right tools and clear instructions, it automates workflows that would typically require hundreds of lines of custom code. This makes it an effective MCP client for both technical and operational tasks.

The Model Context Protocol (MCP) establishes a standard interface for AI to interact with external tools. This makes integration a straightforward configuration change rather than a custom development project. (example)

Relying on natural language also simplifies maintenance. Your internal runbooks effectively become the automation logic. This allows team members to modify and extend workflows without having to navigate complex codebases or proprietary APIs.

Get started

Requirements:

• uv for dependency management
• Claude Code for agent execution

Installation steps:

1. Clone the repository or use it as a template:

   git clone https://github.com/kpolley/easy-agents.git
   cd easy_agents
   

2. Install dependencies:

   uv sync
   

3. Launch the development server:

   uv run fastapi dev main.py
   

Create a custom agent

Follow these three steps:

1. Define the agent's behavior and objectives in natural language.
2. Specify the required MCP servers.
3. Deploy the agent via the FastAPI endpoint or set a cronjob trigger.

Create a custom MCP server

Use FastMCP to build a proprietary MCP server and expand your agent's capabilities. Refer to mcps/github_extended.py for a functional example.