Back to directory

Better Agents

A CLI and set of standards for starting a new agent project, scaffolding scenario tests, versioned prompts and evaluation notebooks so agent behavior stays testable from day one.

Better Agents

Better Agents is a CLI tool and a set of standards for building agents, aimed at making a coding assistant such as Claude Code, Cursor or Kilocode an expert in whichever agent framework a project chooses, such as Agno, Mastra or LangGraph, and in that framework's best practices. Initializing a project generates a standard project structure and an AGENTS.md file that documents conventions the coding assistant then follows, so every new feature comes with scenario tests, versioned prompts and evaluation.

The generated scaffold includes a scenarios directory of agent tests that simulate a conversation to confirm the agent behaves as expected, an evaluations directory of notebooks and datasets for measuring specific prompts on tasks such as RAG or classification, and a prompts directory that stores versioned prompts in YAML, synced through a configuration file for team collaboration and a playground. It also configures MCP servers so the coding assistant understands the chosen framework and can write scenario tests for it. It installs through npm or npx and is written in TypeScript under the MIT license.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.