<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Mateus Trajano</title>
    <description>The latest articles on DEV Community by Mateus Trajano (@mateus_trajano_bafa18af12).</description>
    <link>https://dev.to/mateus_trajano_bafa18af12</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4225115%2Ff741c685-46ff-4642-a864-176bd592782d.png</url>
      <title>DEV Community: Mateus Trajano</title>
      <link>https://dev.to/mateus_trajano_bafa18af12</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/mateus_trajano_bafa18af12"/>
    <language>en</language>
    <item>
      <title>I built Ablit Code: a desktop AI agent that finishes the task instead of chatting about it</title>
      <dc:creator>Mateus Trajano</dc:creator>
      <pubDate>Tue, 03 Nov 2026 23:49:52 +0000</pubDate>
      <link>https://dev.to/mateus_trajano_bafa18af12/i-built-ablit-code-a-desktop-ai-agent-that-finishes-the-task-instead-of-chatting-about-it-35og</link>
      <guid>https://dev.to/mateus_trajano_bafa18af12/i-built-ablit-code-a-desktop-ai-agent-that-finishes-the-task-instead-of-chatting-about-it-35og</guid>
      <description>&lt;p&gt;Most AI coding tools still leave the last mile to you: they suggest, you copy, paste, run, fix and repeat. I wanted an agent that takes the task, does it on my own machine and checks its own work, without an account, without tracking and without a filter deciding what I'm allowed to ask. So I built &lt;strong&gt;Ablit Code&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  What makes Ablit different
&lt;/h2&gt;

&lt;h3&gt;
  
  
  🔓 Abliterated models, built in (coming soon)
&lt;/h3&gt;

&lt;p&gt;Ablit will ship &lt;strong&gt;abliterated (uncensored) open models by default&lt;/strong&gt;: GLM, DeepSeek and others with the refusal layer removed, ready to use with no setup and &lt;strong&gt;no account&lt;/strong&gt;. Monthly plans or pay as you go. Today you can already plug in any provider you like.&lt;/p&gt;

&lt;h3&gt;
  
  
  🔒 Privacy first, for real
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;No account.&lt;/strong&gt; Not for the app, not even to buy models.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No analytics, no telemetry, no tracking.&lt;/strong&gt; Nothing is collected about you.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Your files, chats and settings stay on your computer.&lt;/strong&gt; The only thing that leaves is what you send to the model you chose.&lt;/li&gt;
&lt;li&gt;Optional &lt;strong&gt;PIN lock&lt;/strong&gt; for the app.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  🤖 An agent that finishes the job
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Takes the whole task:&lt;/strong&gt; plans the steps, writes and edits files, runs shell commands, browses the web, then &lt;strong&gt;checks its own result&lt;/strong&gt; before saying it's done.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;40+ built-in tools:&lt;/strong&gt; files, shell, browser, search, HTTP, Python, processes, clipboard, memory, planning and more.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Background jobs:&lt;/strong&gt; long builds, tests and installs run in the background while it keeps working, and it wakes up when they finish.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multi-agent teams:&lt;/strong&gt; split a big job across several agents working at the same time.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;You stay in control:&lt;/strong&gt; risky commands wait for your approval.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  🔌 Any model, any provider
&lt;/h3&gt;

&lt;p&gt;Bring your own keys from any provider. &lt;strong&gt;A free NVIDIA key is enough to start&lt;/strong&gt;, so the whole app costs $0.&lt;/p&gt;

&lt;h3&gt;
  
  
  💻 Everywhere, always up to date
&lt;/h3&gt;

&lt;p&gt;Windows, macOS (Apple Silicon and Intel) and Linux, and it &lt;strong&gt;updates itself&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  How it compares
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;&lt;strong&gt;Ablit Code&lt;/strong&gt;&lt;/th&gt;
&lt;th&gt;Claude Code&lt;/th&gt;
&lt;th&gt;OpenCode&lt;/th&gt;
&lt;th&gt;Cursor&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Account needed&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;No&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Yes (Claude or cloud account)&lt;/td&gt;
&lt;td&gt;No (free models work without one)&lt;/td&gt;
&lt;td&gt;Yes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Analytics / telemetry&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;None&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Usage metrics on by default (opt-out)&lt;/td&gt;
&lt;td&gt;Says none; stores data only on its free models&lt;/td&gt;
&lt;td&gt;Yes (privacy mode optional)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Models&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;Any provider&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Claude only (Anthropic or cloud)&lt;/td&gt;
&lt;td&gt;Any provider&lt;/td&gt;
&lt;td&gt;Cursor's model list&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Uncensored (abliterated) models built in&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;Yes, coming soon&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No (bring your own)&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Price to start&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;Free&lt;/strong&gt; (free NVIDIA key works)&lt;/td&gt;
&lt;td&gt;Paid plan or API&lt;/td&gt;
&lt;td&gt;Free (includes free models)&lt;/td&gt;
&lt;td&gt;Free tier, then paid&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Interface&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;Desktop app&lt;/strong&gt; (Win/Mac/Linux)&lt;/td&gt;
&lt;td&gt;Terminal, IDE, desktop, web&lt;/td&gt;
&lt;td&gt;Terminal, desktop, IDE, web&lt;/td&gt;
&lt;td&gt;Code editor, CLI, web agents&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Background jobs + multi-agent teams&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;Yes&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Yes&lt;/td&gt;
&lt;td&gt;Parallel sessions&lt;/td&gt;
&lt;td&gt;Yes (cloud agents on paid plans)&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;em&gt;Based on each product's public docs as of 3 November 2026; features change, so check their sites for the latest.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Website (and the launch list for the built-in models): &lt;a href="https://ablit-site.ablit.workers.dev" rel="noopener noreferrer"&gt;https://ablit-site.ablit.workers.dev&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Download: &lt;a href="https://github.com/quantumstudent190-netizen/ablit-releases/releases/latest" rel="noopener noreferrer"&gt;https://github.com/quantumstudent190-netizen/ablit-releases/releases/latest&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;It's free. Tell me what breaks, what's confusing and what you'd want it to do next. I read every comment.&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>ai</category>
      <category>productivity</category>
      <category>programming</category>
    </item>
  </channel>
</rss>
