DEDUPED Hot AI repos, turned into use cases Get the Monday 5
180 stars · TypeScript

How to use Gear: what you can do with it

What is Gear? Gear is an open-source framework that iteratively improves an AI agent's harness (instructions, tools, skills, and workflows) using evaluation results from a benchmark of tasks.

How do I use it? Run npm install --global rsi-gear@latest agent-hitch@latest @deepseek-ai/dsh@latest to install Gear, the Hitch evaluation runner, and DSH; then run hitch eval setup harbor to set up the Harbor evaluation runner; install the bundled Refine Skill in your agent environment by following the quick start guide, then configure AutomationBench as the benchmark. Full steps below.

Trended earlier

View repo on GitHub

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer

What do you want to do?

3 use cases

  • Tune a marketing agent on AutomationBench tasks

    Run one Gear optimization round on AutomationBench Marketing tasks to improve an agent's pass rate.

    View prompt
    Use the Gear repo at rsi-gear/gear to run one Refine Skill optimization round on AutomationBench's Marketing tasks for <target agent setup>. Report the pass rate before and after the round and the path to the recorded harness versions. Verify that each retained harness version has a matching evaluation result before you report success.

  • Optimize an agent against your own task set

    Package your own tasks in Harbor format and run Gear's Refine Skill to improve the agent on them.

    View prompt
    Use the Gear repo at rsi-gear/gear to optimize the harness of <target agent> on my task set at <task path in Harbor format>. Report the pass rate for the starting harness and for the best retained harness. Verify that the task paths resolve and that the Refine Skill connects to Gear before the first run.

  • Set a weighted objective that balances pass rate and cost

    Combine pass rate, process scores, and cost or tokens into one objective when tuning a harness.

    View prompt
    Use the Gear repo at rsi-gear/gear to define a weighted objective that combines pass rate and cost for <target task set> and run one optimization round with it. Report the weights used and the pass rate and cost of the retained harness against the starting harness. Verify that the original measurements are still present in the results.

Quick start with Gear

  1. Check you have what it needs: npm installed, to run the global install of rsi-gear, agent-hitch, and DSH; An agent environment such as Codex, Claude Code, or DSH that can use the bundled Refine Skill; A benchmark of representative tasks with clear success criteria, either an existing one or your own in Harbor format.
  2. Open rsi-gear/gear on GitHub and read the README and license.
  3. Run npm install --global rsi-gear@latest agent-hitch@latest @deepseek-ai/dsh@latest to install Gear, the Hitch evaluation runner, and DSH.
  4. Run hitch eval setup harbor to set up the Harbor evaluation runner.
  5. Install the bundled Refine Skill in your agent environment by following the quick start guide, then configure AutomationBench as the benchmark.
  6. In your agent session, invoke /refine or submit a request to improve performance on AutomationBench's Marketing tasks and run one round of optimization.

These steps come from our top use case, Tune a marketing agent on AutomationBench tasks, which also has a copy-paste agent prompt.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for rsi-gear/gear

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Gear alternatives · Best AI repos for Software and DevOps

About this repo

  • Who it is forEngineers and agent builders who have a set of tasks with clear success criteria and want to tune an agent's harness against them.
  • Why it is hot180 stars, launched 69 days ago
  • LanguageTypeScript
  • Sourcegithub.com/rsi-gear/gear · Back to Hot repos

Requirements

  • npm installed, to run the global install of rsi-gear, agent-hitch, and DSH
  • An agent environment such as Codex, Claude Code, or DSH that can use the bundled Refine Skill
  • A benchmark of representative tasks with clear success criteria, either an existing one or your own in Harbor format
  • Access to the models you choose for the Meta agent and the task agent

Add this badge to your README

Maintain Gear? Show readers where to find these use cases. Use cases on Deduped

Markdown 

[![Use cases on Deduped](https://dedupedai.com/badge/rsi-gear/gear.svg)](https://dedupedai.com/use-cases/rsi-gear-gear/?utm_source=github&utm_medium=badge&utm_campaign=rsi-gear-gear)

HTML 

<a href="https://dedupedai.com/use-cases/rsi-gear-gear/?utm_source=github&amp;utm_medium=badge&amp;utm_campaign=rsi-gear-gear"><img src="https://dedupedai.com/badge/rsi-gear/gear.svg" alt="Use cases on Deduped" height="20"></a>

A static image, no tracking script. About the badge

Questions about Gear

What is Gear?

Gear is an open-source framework that iteratively improves an AI agent's harness (instructions, tools, skills, and workflows) using evaluation results from a benchmark of tasks. It is an open-source project on GitHub: https://github.com/rsi-gear/gear.

How do I use Gear?

Run npm install --global rsi-gear@latest agent-hitch@latest @deepseek-ai/dsh@latest to install Gear, the Hitch evaluation runner, and DSH; then run hitch eval setup harbor to set up the Harbor evaluation runner; install the bundled Refine Skill in your agent environment by following the quick start guide, then configure AutomationBench as the benchmark. The use cases on this page give the full steps and a copy-paste agent prompt for each job.

Who is Gear for?

Engineers and agent builders who have a set of tasks with clear success criteria and want to tune an agent's harness against them.

What can I do with Gear?

3 use cases so far, for example: Tune a marketing agent on AutomationBench tasks; Optimize an agent against your own task set; Set a weighted objective that balances pass rate and cost. Each has steps and a copy-paste agent prompt.

What do I need to run Gear?

npm installed, to run the global install of rsi-gear, agent-hitch, and DSH; An agent environment such as Codex, Claude Code, or DSH that can use the bundled Refine Skill; A benchmark of representative tasks with clear success criteria, either an existing one or your own in Harbor format; Access to the models you choose for the Meta agent and the task agent.

What are alternatives to Gear?

Other hot AI repos in Software and DevOps: Deep Infer, Harness2, Orca, Opencode, Graphify.

Is Gear safe to run?

Deduped does not own, audit or vouch for it. Read the README and license, run it on your own systems first and review every command before you run it.