Kilo Code
The open-source agentic engineering platform
Start new thread

Kilo Code v7 for VS Code - Parallel agents, diff reviewer, and multi-model comparisons

by•
We've completely rebuilt Kilo Code for VS Code, built on OpenCode server. New portable core, parallel tool calls, subagent delegation, inline code review, multi-model comparisons. Get started: kilo.ai/install

Add a comment

Replies

Best
Maker
📌

This is the most ambitious release by the team to date.

Looking forward to seeing what you're building with this new extension!

Maker

 thank you, what's your favorite new feature in this release?

Maker

 multi-modal comparisons are so helpful. there's an always-on debate on what the best AI models for coding are. now we know.

[1]:

Maker

 yeah that's very helpful to benchmark performance across different tasks

Maker

 what should we launch next? 👀

Maker
📌

Hey Product Hunt 👋 Brian from Kilo Code here.

We just shipped the biggest update to our VS Code extension since launch. The entire thing has been rebuilt on OpenCode server, which is the same open-source core that powers our CLI and Cloud Agents. One engine across every surface, so when we improve something, it gets better everywhere.

The headline feature is real parallelism. Kilo can now run multiple tool calls at the same time (file reads, searches, terminal commands all firing concurrently), and it can spin up parallel subagents that each handle a piece of a larger task simultaneously. You actually feel the speed difference.

A few other things shipping in this release:

  • Agent Manager — run multiple independent agents in separate tabs, give each one a role, and use git worktrees so they never step on each other's code

  • Inline code review — leave line-level comments directly on agent diffs, just like a PR review, and send them back as structured context

  • Multi-model comparisons — run the same prompt through different models side by side and pick the best result

  • Cross-platform sessions — start in the CLI, pick up in VS Code, share with a teammate

Kilo is open source, runs 500+ models at provider cost (zero markup), and has over a million developers using it. We'd love for you to try it out and tell us what you think!

Maker

 agent manager is such a great addition, and overall it works so much faster! Great to be working together on this one!

Maker

   great product. great team. period.

Maker
Maker

 curious what's your look like? any extensions you'd recommend in addition to ?

Maker

 I added a front-end specialist skill from , but other than that - the vanilla extension is all I need for my work!

Maker

 KISS ftw

 One engine across CLI, VS Code, and Cloud Agents is the right architecture. We run agent workflows across multiple surfaces internally and the biggest friction is always inconsistency between environments. Something works in the terminal, behaves differently in the editor, breaks entirely in cloud. A unified core that improves everywhere simultaneously solves that at the infrastructure level.

Real parallelism is the feature that matters most here. Sequential tool calls are the hidden bottleneck in most agentic coding workflows. When a complex task requires reading 10 files, searching a codebase, and running terminal commands, doing that one at a time turns a 30-second task into a 5-minute wait. Parallel subagents handling pieces of a larger task simultaneously is where the speed compounds even further, especially on architectural work that naturally decomposes into independent subtasks.

The Agent Manager with git worktrees is a smart detail. We've run into the exact problem of multiple agents stepping on each other's code. Isolating each agent in its own worktree so they can operate independently without merge conflicts is the kind of practical engineering decision that shows you've actually dealt with multi-agent workflows in production, not just theorized about them.

Inline code review on agent diffs is great too. The gap between "agent generates code" and "human approves code" is usually copy-paste into a PR tool. Doing it directly in the editor with structured context going back to the agent tightens that feedback loop significantly.

Open source at provider cost with zero markup across 500+ models. Hard to argue with that. Congrats on the release.

Maker

 Thank you very much for the thoughtful comment!

 Of course! Been following your products' development since you guys started. Big fan.

Maker

  is the 🐐 - thanks for the continuous support!

 Hey Brian, what stops two subagents from making contradictory edits to the same file when they're working on related parts of a task? You mentioned git worktrees for the Agent Manager; is the same isolation used for subagents, or is there a different mechanism?

  congratulations on the release! I am new to Kilo Code but am looking forward to trying it out. I commented under post because I am also curious about this, and would you still "recommend" agents work on separate tasks or even separate files when possible to avoid conflicts or is the Agent Manager able to handle?

Maker

 Exactly - it's git worktrees for Agent Manager, and with subagents it depends - if the tool calls won't conflict or aren't actually writing code (such as read or plan agents), worktrees aren't required.

Maker

    I'd just add that subagents are delegated by a parent agent mainly to keep its context focused - they go do the work, and return a summary to the parent rather than the full tool-call history. They share the parent's workspace rather than getting their own worktree, so the safeguard against contradictory edits there is the parent orchestrating: subagents are best used for scoped exploration and well-bounded changes, while the parent integrates.

Where Agent Manager really shines is when the tasks are already structurally independent (e.g., different features, different files, or "throw the same prompt at three models and pick the best result") - it gives you one surface to track all of them, see PR/review status, and prune what you don't want.

More on that in our livestream that aired earlier today and is now available on demand. Marius, the engineer behind Agent Manager, breaks it down here (jump to 18:00)

Been on Cursor for a while but the model lock-in is starting to bother me. The 500+ models angle here is hard to ignore. Does switching models mid-project break any context?

Maker

 happy to hear, and no, you can even switch model mid task.

Maker

exactly - 500+ models, zero markup, no editor switch required. see full comparison vs Cursor here:

hope this helps!

Maker

anything else should build/improve/fix from your perspective?

Maker

 It does not! You can switch in between prompts, or even run parallel agents using different models to see how they handle the tasks differently.

Your context remains intact!

Maker

   curious what are your preferred AI models when coding?

Is the ability to switch between models the main reason to try this? I’m a daily user or Claude code and codex. Tried antigravity and gave up. Planning to try KimiCode. 😅 Should I add kilo code to the list?
Maker

definitely! read this full breakdown:

TL,DR: is open-source, offers multi-model CLI + IDE agent with inline autocomplete vs Anthropic's Claude-only terminal-first coding agent.

hope it helps!

Maker

 i might be biased, but you definitely should ;) - you can also use your codex subscription inside of Kilo, or use any of the free/frontier models

Maker

 oss ftw!

Maker

 I certainly think so!

In addition to freely switching models, we're also completely open-source and focused on frictionless agentic orchestration. The openness combined with parallel execution makes it a different experience than other tools, in my humble opinion!

Maker

View source code on GitHub:

Star this repo!

will add to the list. 😅
Maker

awesome! feel free to join the Discord server for advanced support:

looking forward to your thoughts

 same boat — claude code daily here. model switching alone isn't the hook (i already pick opus/sonnet/haiku per task). what i'd actually try kilo for is parallel subagents on git worktrees — still scripting that by hand today and the merge collisions get messy fast.

Maker

awesome! open-source, multi-model CLI + IDE with inline autocomplete... fwiw you can find a

enjoy!

The inline diff reviewer is the feature I didn't know I needed. Been using AI coding tools for a year and reviewing agent changes is still the most painful part of the workflow. Does it support split views on smaller screens? Congrats BTW )

 thank you, great to hear you like the inline diff reviewer. I use the code reviewer in VS Code before creating a PR, and let the code reviewer in the cloud check everything again. For small changes (and screens, since I do a lot of that on mobile) I usually kick if off from Slack or cloud agents, and then have the code reviewer in the cloud do the check.

Maker

 yes! and I'd add you can learn more about Kilo Code Reviewer (launched on Product Hunt ) here:

Maker

The inline diff reviewer is the feature I didn't know I needed.

love it! give it a spin at and let us know how it goes with a review here:

looking forward to it!

Maker

what should launch nextfrom your perspective? take the survey here in

Maker

 Thank you!

Yes it does still support split views - even with multiple agents running!

Maker

 any AI model you'd recommend for code reviews btw?

Came from Roo Code a few months ago and honestly haven't looked back. The migration was smoother than expected. Good luck for today!

Maker

 happy to hear you like it. Have you tried agent manager already to put multiple agents to work?

Maker

also curious how many coding agents do you run in parallel?

Maker

Came from Roo Code a few months ago and honestly haven't looked back. The migration was smoother than expected.

Your words just made our day, Olivia!

For anyone else reading this and coming over from Roo Code, the team wanted to make this as easy as possible. They crafted a migration guide that walks through bringing settings, modes, and workflows into Kilo here:

Enjoy! and make sure to leave a review here:

Maker

oh and one more thing: what should launch next? 👀

Maker

 Love to hear that! Thank you!

Maker

   spreading the word on X!

Been using v7 for a while, and even the versions before this. Took a while to get used to, but generally excited to see it being built on OpenCode. Can't wait for more features especially /remote 🤞🙏

Maker

 love to read this, Lim! make sure to and help us 💛🖤

Maker

 also curious what's your look like? any extensions you'd recommend in addition to ?

Maker

 thank you for using Kilo!

Maker

   also what else should build/improve/fix? here to build the best DX as possible

Maker

 Thank you! Glad you're vibing with it ;)

Maker

  do you have a remote feature in the backlog? a target date? #featurerequest cc

If one agent is refactoring an API and another agent is consuming that API, how do they handle the dependancy? Do they share context live or waiting for human reviews?

Maker

 you can use agent manager to let agents run on separate git worktrees so they don't interfere with each other. You can then review locally and in the cloud with Kilo, and resolve any merge conflicts.

Maker

FYI you can read the docs for full details on the Agent Manager here:

Maker

 They use git worktrees to avoid conflict and share context upon merge!

Maker

curious: how do you like to work with AI coding agents? do you like to have a look at the context usage (Human-in-the-Loop) or prefer to review the outputs (agent-first)?

Multi-model comparisons inside the editor is the feature I didn’t know I needed. Does it run them simultaneously or sequentially?
Maker

 they run simultaneously!

Maker

 any preferred AI models when coding btw? heard good things about

Maker

Multi-model comparisons inside the editor is the feature I didn’t know I needed.

framing this!

Maker

People embrace this feature on X, too:

This side-by-side multi-model thing is 🔥 Game changer for real.

Source:

Maker

 it runs them simultaneously on separate worktrees

Maker

 Either!

You can run them in parallel inside the Agent Manager or just start a new branch and run them sequentially.

Maker

re: Agent Manager. FYI you can read the docs for full details on this new feature:

Using Kilo for Resume Matcher. Works really well, however, ƒor this launch. I'd suggest some features that Kilo should add. The first one is Skills, just like Claude-Code. A .kiloignore, and custom routines to be fired up to check for dependencies, security risks, and other supply chain attacks that may be in transitive dependencies. Because the more we vibe-code, the better the security should be, and the more exhaustive the reviews should be.

Maker

 big fan of your work, Saurabh! what do you enjoy the most with ?

appreciate the feedback 💛🖤 cc

   thanks.

Kilo is customizable, both on coding and PR reviews. I'm using it mainly for PR reviews, right now.

Maker

 love it - what's your look like btw? any extensions you'd suggest in addition to ?

 Nope, for reviews Kilo Code is the best. Rest what I use is the Claude for VS Code one.

Maker

 thanks for using Kilo and your feedback. We do support skills, and we also have our marketplace! . We agree security matters! We have been working on a few features in that space: , If you have further feedback please let us know on our GitHub repo!

Maker

direct link to the gh repo icymi:

 Awesome, I'll test the Security part really soon.

Maker

   keep us posted! and make sure to leave a review here:

Maker

 Thanks for the feedback! Both Skills and .kiloignore are currently supported in v7 - nice callout!

Maker

what else should build/improve/fix on your experience?

123
•••
Next
Last