The latest articles on DEV Community by Arnab Karmakar (@arnab_karmakar_41f6ef02ca). https://dev.to/arnab_karmakar_41f6ef02ca 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%2F4034788%2F9c0f7962-c59f-4070-a07c-8966dc6449d5.png https://dev.to/arnab_karmakar_41f6ef02ca en Arnab Karmakar Sat, 18 Jul 2026 06:33:44 +0000 https://dev.to/arnab_karmakar_41f6ef02ca/i-built-an-open-source-workspace-to-tame-my-ai-coding-agents-pb2 https://dev.to/arnab_karmakar_41f6ef02ca/i-built-an-open-source-workspace-to-tame-my-ai-coding-agents-pb2 <h2> The problem </h2> <p>Every AI coding CLI is a snowflake: different flags, different session models, and all of them editing your one working copy. I wanted to run agents in parallel on isolated branches, queue prompts while an agent works, and have a real UI — without handing API keys to yet another app.</p> <h2> What AgentGrove is </h2> <p><a href="https://github.com/arnabk/agentgrove" rel="noopener noreferrer">AgentGrove</a> is an open-source (MIT) local developer workspace. It launches the agent CLIs already on your machine (Claude Code, opencode, Kimi), translates their event streams into one UI, and scopes every chat to a git worktree.</p> <h2> Feature tour </h2> <ul> <li> <strong>Multi-provider agents</strong> — Claude, opencode, Kimi; live model lists, session resume, per-chat model/effort settings</li> <li> <strong>Worktrees everywhere</strong> — each chat can live in its own worktree; create/restore from the UI, pre-scripts for <code>pnpm install</code> et al, celestial branch names with a galaxy map of where you've been</li> <li> <strong>Real chat UX</strong> — streaming with coalescing, thinking blocks, tool activity rail, retry, forking, prompt-revert (undo the files a prompt touched)</li> <li> <strong>Prompt queue</strong> — messages queue while the agent works; auto-drain or manual mode</li> <li> <strong>Postgres DB editor</strong> — saved connections with inline test, table browser with filters + pagination, SQL editor with schema autocomplete</li> <li> <strong>The basics done right</strong> — PTY terminal, CodeMirror editor with autosave, sub-10ms fuzzy file search over 100k files, notes scratchpad, team chat</li> <li> <strong>Rust + SolidJS</strong> — single-binary backend, low memory footprint, hot-reload dev</li> </ul> <h2> How it works </h2> <p>No vendor SDKs. Each provider is a small subprocess adapter (~400 lines of Rust) that launches the CLI with its JSON stream flags and translates events into a common enum (<code>Token</code>, <code>Thinking</code>, <code>ToolCall</code>, <code>ToolResult</code>, <code>Done</code>, <code>Error</code>). Auth stays with your local CLI — the app never sees a key. Adding a provider means writing one translator and registering it.</p> <h2> Testing </h2> <p>Two layers keep this honest: a deterministic fake provider for dispatch/queue/resume e2e tests, and real-model Playwright runs against the actual CLIs. The demo videos on the README are recorded by Playwright in an isolated Docker stack (including a seeded Postgres for the DB editor demo), not staged screenshots.</p> <h2> Try it </h2> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>git clone https://github.com/arnabk/agentgrove.git <span class="nb">cd </span>agentgrove <span class="o">&amp;&amp;</span> just dev </code></pre> </div> <p>Demo videos are on the <a href="https://github.com/arnabk/agentgrove" rel="noopener noreferrer">README</a>. Issues and PRs welcome — I'd love to hear which provider you want next.</p> ai rust opensource devtools