LangChain
Agent engineering platform for building LLM applications from composable parts
Best for: developers who want to study how an autonomous research agent plans, reflects and revises — using financial analysis as the worked example.
Dexter is an autonomous financial research agent built by Virat Singh. It performs analysis through task planning, self-reflection and real-time market data, and the README frames it plainly as "Claude Code, but built specifically for financial research." It can be driven from the command line or through WhatsApp.
What makes it interesting: the agent loop is legible. Planning, self-critique and evaluation are separated clearly enough that you can read how each stage works, which makes it a useful reference implementation if you are building research agents in any domain.
Where it falls short: read the disclaimer before anything else. The author states the project is for educational, entertainment and informational purposes only, is not intended for real trading or investment, and is not financial advice. It also carries no declared license, which is a genuine constraint if you plan to build on it commercially. Treat Dexter as a study of agent architecture, not as a trading system.
Details: TypeScript. No declared license.
Agent engineering platform for building LLM applications from composable parts
Agent harness optimisation for Claude Code, Codex, OpenCode and Cursor
Self-improving agent from Nous Research that learns across sessions
Small, composable agent skills from Matt Pocock's daily .agents directory