An open-source experimental CLI-based coding agent built with CrewAI Flows. This agent is designed to autonomously plan and execute coding tasks, leverage web research, and adapt to your project's conventions within your cli.

• 🧠 Persistent Memory: Stores project conventions and learned information in markdown files (~/.crewai_agent/memories), allowing the agent to "remember" your coding style and preferences across sessions.
• 🌊 CrewAI Flows: Orchestrates a two-stage process:
  1. Planning Crew: Analyzes your request, checks memories, and researches the web (using Exa) to create a detailed implementation plan.
  2. Execution Crew: Autonomously executes the plan using a suite of file and exploration tools.
• 🔍 Web Research: Integrated ExaSearchTool and ScrapeWebsiteTool allow the agent to find the latest documentation and examples for unknown libraries.
• 🛠️ Subagent Capabilities: The Execution agent is equipped with:
  • GrepTool & ListDirectoryTool: To explore and understand the existing codebase.
  • TodoPlanningTool: To self-manage complex tasks by creating and tracking a TODO.md file.
• ⚡ Skills System: Extensible architecture inspired by Claude Skills. Add new capabilities by simply creating a SKILL.md file in a .skills/ directory.

1. Clone the repository:

   git clone <repository-url>
   cd cli_agent

2. Install dependencies: Requires Python 3.10+.

   pip install -e .

You must set the following environment variables for the agent to function:

export GEMINI_API_KEY="your_google_gemini_api_key" # or GOOGLE_API_KEY
export EXA_API_KEY="your_exa_api_key"

Run the agent with a natural language description of your task:

cli-agent start "Create a python script that fetches the latest news about AI using requests and beautifulsoup"

The agent will:

1. Load relevant memories.
2. Research the web if needed.
3. Create a plan.
4. Execute the plan, creating files and writing code.

Teach the agent about your project conventions or preferences:

# Add a new memory
cli-agent memory add conventions "Always use type hints and docstrings in Python code."

# List all memories
cli-agent memory list

The agent can dynamically load "Skills" to perform specialized tasks.

1. Create a Skill: Create a directory .skills/my-skill/ in your project root (or ~/.crewai_agent/skills/). Add a SKILL.md file with instructions:

   ---
   name: my-skill
   description: A brief description of what this skill does.
   ---
   # My Skill Instructions
   1. Do step one.
   2. Do step two.

2. Use the Skill: Ask the agent to use it:

   cli-agent start "Use the my-skill to perform X"

• code-reviewer: A skill to review code for best practices and bugs.
• excalidraw-architect: Generates architectural diagrams in Excalidraw JSON format.

• src/cli_agent/flow.py: The main entry point using CodingAgentFlow.
• src/cli_agent/crews/planning_crew.py: The PlanningCrew definition.
• src/cli_agent/crews/execution_crew.py: The ExecutionCrew definition.
• src/cli_agent/utils/memory_manager.py: Handles reading/writing memory files.
• src/cli_agent/tools/: Custom tools for file operations, search, and skills.

MIT License