Kilo Code Returns to Product Hunt: Agentic Development at Scale

We're back on Product Hunt — here's how far Kilo has come

TLDR: Kilo Code relaunched on Product Hunt with a dramatically expanded product: a parallel-agent VS Code extension, CLI, Cloud Agents, and Slack integration sharing a single engine. The biggest architectural change since launch includes an Agent Manager, git worktrees, and 2.3 million installs.

Summary: A year ago, Kilo Code was a VS Code extension with a chat window. Today it's something quite different, and the relaunch on Product Hunt is as much a progress report as a product launch. I find this kind of "we're back and here's what changed" post genuinely useful because you can measure the distance between ambition and delivery. In Kilo's case, the distance is real.

The core shift is from "AI assistant in your editor" to "runtime for multiple agents running in parallel." That sounds like marketing, but the mechanics are concrete. The new VS Code extension has an Agent Manager that lets you open multiple Kilo tabs simultaneously, assign each one a role, and run them on isolated git worktrees so they don't step on each other. One agent writes the API endpoint, another handles the auth refactor, a third runs the test suite. You review the diffs when they're done. That's a fundamentally different model than autocomplete or even single-agent chat.

The parallelism goes deeper than the Agent Manager. Within a single agent session, file reads, searches, and terminal commands now execute concurrently. That matters because the slowest part of agentic coding is usually the tool calls, and sequential tool execution turns a 10-second task into a 90-second one. The speed difference when you remove that bottleneck is something you feel immediately, not just in benchmarks.

What I respect about this post is the section on the GA release. They shipped a major architectural rewrite to their full install base, it broke things at scale that the pre-release couldn't surface, and they ran three weeks of public iteration with 188 PRs in week two alone. Memory spikes on Windows, rate-limit edge cases, session stability problems. They fixed them in public with weekly updates. That's a hard thing to do and a harder thing to write about honestly.

The model comparison feature is interesting too. You can run the same prompt against Claude, GPT, Gemini, and any of 500+ other models simultaneously and compare results side by side. For architecture decisions or open-ended refactors where you genuinely don't know which model will give you the best answer, that's a practical tool, not a gimmick.

Key takeaways:

• Agent Manager enables true parallel agentic work with git worktree isolation per agent
• The new engine is shared across VS Code, CLI, Cloud Agents, and Slack for session continuity
• The GA rollout was rocky but transparent, with 188 PRs in week two fixing real production issues
• Kilo Code remains free and open source, with BYOK support and no model lock-in

Why do I care: Parallel agents with worktree isolation is the architecture I've been waiting for. The ability to delegate a module refactor, an API endpoint, and test generation to three separate agents running simultaneously, then review structured diffs, changes how you think about what "a day of coding" looks like. The open-source, BYOK model also matters: I don't want to be locked into a vendor's model choices when the model landscape is shifting this fast.

We're back on Product Hunt — here's how far Kilo has come