<?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: Madhubalan Appachi</title>
    <description>The latest articles on DEV Community by Madhubalan Appachi (@madhuappachi).</description>
    <link>https://dev.to/madhuappachi</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.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3773965%2Fde6201cc-ec6b-4139-b27b-6e38e27b1f9d.jpg</url>
      <title>DEV Community: Madhubalan Appachi</title>
      <link>https://dev.to/madhuappachi</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/madhuappachi"/>
    <language>en</language>
    <item>
      <title>I built a local-only shell history manager in Rust with AI agent tracking</title>
      <dc:creator>Madhubalan Appachi</dc:creator>
      <pubDate>Sat, 14 Mar 2026 13:22:33 +0000</pubDate>
      <link>https://dev.to/madhuappachi/i-built-a-local-only-shell-history-manager-in-rust-with-ai-agent-tracking-fh0</link>
      <guid>https://dev.to/madhuappachi/i-built-a-local-only-shell-history-manager-in-rust-with-ai-agent-tracking-fh0</guid>
      <description>&lt;p&gt;Every shell history tool I tried either wanted me to create a cloud account or had no way to track what my AI agents were running.&lt;/p&gt;

&lt;p&gt;So I built &lt;a href="https://github.com/AppachiTech/suvadu" rel="noopener noreferrer"&gt;Suvadu&lt;/a&gt; — a local-only, SQLite-backed shell history manager with a TUI, agent monitoring, and automatic secrets redaction. No cloud, no account, no data leaving your machine.&lt;/p&gt;

&lt;p&gt;The name is Tamil for "trace" or "footprint."&lt;/p&gt;

&lt;h2&gt;
  
  
  Why I built something different
&lt;/h2&gt;

&lt;p&gt;Atuin is a great tool — and yes, you can run it in local-only mode. But Atuin is designed around sync. The account system, the server component, the encryption layer — that's where the engineering focus is, and that's what makes it powerful for teams and multi-machine setups.&lt;/p&gt;

&lt;p&gt;I wanted something different: a tool that's &lt;strong&gt;local-first by design&lt;/strong&gt;, not local as an opt-out. No server component in the codebase, no account system, no sync protocol — just a SQLite database and a TUI. Fewer moving parts, smaller attack surface, simpler mental model.&lt;/p&gt;

&lt;p&gt;The other thing I couldn't find in any existing tool was &lt;strong&gt;AI agent monitoring&lt;/strong&gt;. With Claude Code, Cursor, and Aider running commands autonomously, I wanted to know exactly what they were executing and flag risky operations (&lt;code&gt;rm -rf&lt;/code&gt;, &lt;code&gt;git push --force&lt;/code&gt;, permission changes) before they cause damage.&lt;/p&gt;

&lt;p&gt;So Suvadu is built around two ideas: &lt;strong&gt;local-only by architecture&lt;/strong&gt; and &lt;strong&gt;AI agents as first-class citizens in your history&lt;/strong&gt;.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Feature&lt;/th&gt;
&lt;th&gt;Suvadu&lt;/th&gt;
&lt;th&gt;Atuin&lt;/th&gt;
&lt;th&gt;McFly&lt;/th&gt;
&lt;th&gt;Hstr&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Storage&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Local SQLite&lt;/td&gt;
&lt;td&gt;Cloud or local&lt;/td&gt;
&lt;td&gt;Local SQLite&lt;/td&gt;
&lt;td&gt;In-memory&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Cloud sync&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;No (local-only by design)&lt;/td&gt;
&lt;td&gt;Yes (local mode available)&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;TUI&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Full-screen + detail pane&lt;/td&gt;
&lt;td&gt;Full-screen&lt;/td&gt;
&lt;td&gt;Inline overlay&lt;/td&gt;
&lt;td&gt;Full-screen&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;AI agent tracking&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Auto-detect + risk assess&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Secrets redaction&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Auto before storage&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Activity heatmap&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Yes (5-tier)&lt;/td&gt;
&lt;td&gt;Dashboard (cloud)&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Themes&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;3 built-in, hot-swap&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;td&gt;No&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Search
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fkzhvgrkiuht8vawpou92.gif" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fkzhvgrkiuht8vawpou92.gif" alt="Suvadu search TUI showing fuzzy search with detail pane, unique mode toggle, tag filtering, and bookmarks" width="800" height="457"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Fuzzy search powered by &lt;a href="https://github.com/helix-editor/nucleo" rel="noopener noreferrer"&gt;nucleo&lt;/a&gt; (same engine as Helix editor). Features:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Toggleable detail pane showing full metadata for each entry&lt;/li&gt;
&lt;li&gt;Smart mode — context-aware ranking that boosts same-directory commands&lt;/li&gt;
&lt;li&gt;Directory scoping, date filters, tag filters, executor filters&lt;/li&gt;
&lt;li&gt;Unique-command deduplication toggle&lt;/li&gt;
&lt;li&gt;Bookmarks and notes on any entry&lt;/li&gt;
&lt;li&gt;Responsive layout that adapts to terminal width&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Stats
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fxk458dtk0ivq0do4io92.gif" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fxk458dtk0ivq0do4io92.gif" alt="Suvadu stats dashboard with GitHub-style activity heatmap cycling through 7-day, 30-day, and yearly views with executor breakdown" width="1400" height="800"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;A GitHub-style activity heatmap with 5 intensity tiers. Cycle through time periods (30d / 90d / 180d / 365d) with a single keypress. Includes an executor breakdown showing which tools ran what.&lt;/p&gt;

&lt;h2&gt;
  
  
  Agent monitoring
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fmgxhiqmkauyci8wcucpm.gif" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fmgxhiqmkauyci8wcucpm.gif" alt="Suvadu agent dashboard showing AI command timeline with risk indicators, agent filtering between Claude Code and Cursor, and risk-level filtering" width="800" height="457"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;This is the feature I use the most. Every command run by an AI agent gets automatically detected and risk-assessed:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Critical&lt;/strong&gt;: &lt;code&gt;rm -rf /&lt;/code&gt;, &lt;code&gt;git push --force origin main&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;High&lt;/strong&gt;: &lt;code&gt;chmod 777&lt;/code&gt;, &lt;code&gt;npm install&lt;/code&gt;, config overwrites&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Medium&lt;/strong&gt;: &lt;code&gt;git reset&lt;/code&gt;, &lt;code&gt;docker run&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Safe&lt;/strong&gt;: &lt;code&gt;git status&lt;/code&gt;, &lt;code&gt;ls&lt;/code&gt;, &lt;code&gt;cargo test&lt;/code&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;There's an interactive dashboard with a timeline, per-agent analytics, and exportable reports (text, markdown, JSON). You can filter by agent, risk level, or time period.&lt;/p&gt;

&lt;p&gt;Setting it up for Claude Code is one command:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;suv init claude-code
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Privacy and security
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Commands prefixed with a space are never recorded&lt;/li&gt;
&lt;li&gt;Configurable regex exclusion patterns&lt;/li&gt;
&lt;li&gt;API keys, tokens, and passwords are auto-redacted before they hit the database&lt;/li&gt;
&lt;li&gt;Database and config files use owner-only permissions (0o600)&lt;/li&gt;
&lt;li&gt;Self-update verifies binary signatures via Minisign&lt;/li&gt;
&lt;li&gt;All data stays local — no telemetry, no external servers, ever&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Other bits
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Session timeline&lt;/strong&gt; — browse history by terminal session with an interactive TUI&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Alias suggestions&lt;/strong&gt; — analyzes your history and suggests shell aliases for frequently-typed commands&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Export/import&lt;/strong&gt; — JSONL, CSV, JSON, and &lt;code&gt;~/.zsh_history&lt;/code&gt; import&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Shell completions&lt;/strong&gt; — Zsh, Bash, and Fish&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  The stack
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Rust&lt;/strong&gt; with &lt;a href="https://ratatui.rs" rel="noopener noreferrer"&gt;ratatui&lt;/a&gt; for the TUI&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;SQLite&lt;/strong&gt; (via rusqlite) with WAL mode for storage&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;nucleo-matcher&lt;/strong&gt; for fuzzy search&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Minisign&lt;/strong&gt; for release signature verification&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Zsh and Bash hooks&lt;/strong&gt; for transparent recording&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Install
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# macOS (Homebrew)&lt;/span&gt;
brew tap AppachiTech/suvadu &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; brew &lt;span class="nb"&gt;install &lt;/span&gt;suvadu

&lt;span class="c"&gt;# macOS / Linux (direct download)&lt;/span&gt;
curl &lt;span class="nt"&gt;-fsSL&lt;/span&gt; https://downloads.appachi.tech/install.sh | bash

&lt;span class="c"&gt;# From source&lt;/span&gt;
cargo &lt;span class="nb"&gt;install &lt;/span&gt;suvadu
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then add to your shell config:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;eval&lt;/span&gt; &lt;span class="s2"&gt;"&lt;/span&gt;&lt;span class="si"&gt;$(&lt;/span&gt;suv init zsh&lt;span class="si"&gt;)&lt;/span&gt;&lt;span class="s2"&gt;"&lt;/span&gt;   &lt;span class="c"&gt;# or bash&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Links
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub&lt;/strong&gt;: &lt;a href="https://github.com/AppachiTech/suvadu" rel="noopener noreferrer"&gt;github.com/AppachiTech/suvadu&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Website&lt;/strong&gt;: &lt;a href="https://www.appachi.tech/suvadu" rel="noopener noreferrer"&gt;appachi.tech/suvadu&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;It's MIT licensed and I'd welcome feedback — especially if you're using AI coding agents and want better visibility into what they're running. Star the repo if it's useful.&lt;/p&gt;

</description>
      <category>rust</category>
      <category>terminal</category>
      <category>cli</category>
      <category>opensource</category>
    </item>
  </channel>
</rss>
