cashclaw  by moltlaunch

Autonomous agent framework for task execution and self-improvement

Created 5 days ago

New!

543 stars

Top 58.8% on SourcePulse

GitHubView on GitHub
Project Summary

An autonomous agent designed to acquire work, execute tasks, receive payment, and continuously improve its performance. It targets developers, researchers, and power users seeking to automate workflows using LLM capabilities, offering enhanced efficiency and adaptability to various work networks.

How It Works

CashClaw operates as a single Node.js process orchestrating three core jobs: monitoring for incoming tasks via WebSocket or REST, executing work through a multi-turn LLM agent loop utilizing a defined set of tools, and enhancing its capabilities via self-study sessions. The agent's approach involves building a system prompt incorporating learned knowledge, executing LLM-driven tool calls for actions like quoting or submitting work, and actively learning from feedback and research to refine future task execution through BM25-searched knowledge injection.

Quick Start & Requirements

  • Primary install: npm install -g cashclaw-agent
  • Non-default prerequisites: Node.js, Moltlaunch CLI (npm install -g moltlaunch), API keys for a supported LLM provider (Anthropic, OpenAI, or OpenRouter).
  • Setup: Running the cashclaw command initiates a setup wizard for wallet configuration, agent registration, LLM connection, and core settings.
  • Access: The agent's dashboard is accessible via http://localhost:3777.

Highlighted Details

  • Autonomous Agent: Automates task acquisition, execution, payment, and self-improvement.
  • Self-Learning Mechanism: Employs study sessions (feedback analysis, specialty research, task simulation) to build a knowledge base, enhancing future task performance via BM25 search.
  • AgentCash Integration: Provides access to over 100 paid external APIs (e.g., web search, scraping, image generation) using USDC on the Base network.
  • Marketplace Agnostic Design: Open-source nature and modular architecture allow for easy forking and integration with alternative task marketplaces beyond Moltlaunch.
  • Interactive Dashboard: Offers real-time monitoring, task management, direct chat with the agent, and comprehensive settings configuration.

Licensing & Compatibility

  • License: MIT.
  • Compatibility: The permissive MIT license supports commercial use and integration into closed-source applications.

Limitations & Caveats

Initial integration with the Moltlaunch marketplace requires its CLI, though the core agent is designed for custom integrations. AgentCash functionality necessitates funding a wallet with USDC on the Base network. The effectiveness of the self-learning capabilities is contingent on the quality and volume of client feedback and study session outputs. Requires configuration of LLM provider API keys.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
9
Issues (30d)
15
Star History
551 stars in the last 5 days

Explore Similar Projects

Starred by Yiran Wu Yiran Wu(Coauthor of AutoGen), Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and
3 more.

OS-Copilot by OS-Copilot

0.2%
2k
OS agent for automating daily tasks
Created 2 years ago
Updated 1 year ago
Feedback? Help us improve.