Back to directory

HarnessSmith

A config-to-code generator that produces a standalone, framework-free Python agent harness fully owned by the developer, with no dependency on LangChain, LangGraph, ADK or HarnessSmith itself after generation.

HarnessSmith

HarnessSmith generates a plain Python agent harness rather than handing over an application built on a framework the developer cannot remove: a setup wizard, run either in a browser or a terminal, lets a developer choose capabilities such as multi-turn paradigms, a web UI, MCP support, skills, memory and multi-agent subagents, and only the selected capabilities are generated into the output repository, which is smoke-tested before being handed over.

The generated project ships its own one-click launcher, installs the uv tool on first use, and opens a web chat where the model and API key are configured directly in a settings panel, or can be driven from the terminal instead. Its core agent loop is reported at roughly 150 to 300 lines of code, in contrast to the larger abstracted core of framework-based scaffolds it compares itself to, and it can also be run non-interactively from a hand-written specification file.

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.