mupt-ai

mupt-ai / dari-docs

Public

optimize your documentation through fleets of agents

optimize.dari.dev
34
0
69% credibility
Found May 21, 2026 at 34 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Go
AI Summary

dari-docs is a documentation quality testing tool that evaluates whether written instructions are clear enough for AI agents to follow. It works by having simulated AI 'testers' attempt real tasks using only your documentation, then reports where ambiguity, missing steps, or unclear terminology blocked their progress. The tool supports two modes: a quick 'check' that generates feedback reports, and an 'optimize' mode that also proposes specific fixes to your docs. Users can run it locally against their own AI agents or use a hosted service that handles everything automatically. The project includes a web interface for managing runs, authentication via browser login or API keys, and integrates with payment systems for the hosted version.

How It Works

1
📝 You have documentation to test

You're a developer or tech writer who wants to know if your docs are actually clear enough for someone (or something) to follow.

2
🔧 You install the tool in one line

A simple install script gets the tool running on your computer, ready to test any docs folder you point it at.

3
🎯 You give it a task to try

Instead of hoping someone reads your docs, you describe a real goal like 'install the SDK and make your first API call' and let AI agents attempt it.

4
🤖 AI agents attempt the task using only your docs

Simulated developers try to complete your task, reading your documentation exactly as someone new would, and trying the steps themselves.

5
You see the results
📋
Just checking

See the feedback report and fix your docs yourself, with concrete evidence of what needs improvement.

✏️
Get help fixing

Let an editor agent propose actual changes to your docs, which you can review and accept before applying.

6
✅ Your docs are now agent-ready

With feedback in hand (and optionally auto-generated fixes), your documentation becomes something AI can actually follow and use.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 34 to 34 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is dari-docs?

dari-docs is a Go-based CLI that stress-tests your documentation by throwing it at AI agents and watching where they get stuck. You point it at a docs folder or URL, give it a task like "install the SDK and make a first API call," and it spins up simulated developers who try to complete that task using only your docs. The tool then reports back exactly where ambiguity, missing setup steps, or unclear terminology blocked progress. The `optimize` command goes further--it generates proposed documentation edits based on that feedback, which you can review before merging. It runs in two modes: managed (uses the hosted dari.dev service with $5 free credits for new accounts) or self-managed (run against agents in your own dari.dev organization).

Why is it gaining traction?

The premise is clever: if an AI agent can complete a task from your docs, a human probably can too. But the real hook is the feedback loop. Instead of guessing whether your docs are clear, you get concrete failure reports with specific locations where agents had to guess or infer. The `optimize` command then turns that pain point data into actual markdown edits you can review. It's a practical answer to the question every docs team asks: "how do we know if this is actually good?"

Who should use this?

Technical writers maintaining SDK or API docs who want objective feedback beyond "I found it confusing." DevRel teams shipping documentation alongside developer products. Open source maintainers who want to verify their README actually helps new contributors get started. Not for marketing docs or internal wikis--this is specifically for documentation that developers need to complete tasks.

Verdict

A genuinely interesting idea with a clean CLI interface and solid Go implementation. At 34 stars, it's early-stage and the credibility score of 0.7% reflects that--test coverage and polish will improve with adoption. Worth trying on a small docs project to see if the feedback loop adds value to your review process. The hosted option lowers the barrier to experimentation.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.