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
Very excited for this launch. The new Kilo for VS Code is my daily driver, and this rebuilt version with agent manager is in my opinion the next step in agentic coding. I now let multiple agents run at the same time using agent manager, and it speeds up my workflow a lot. Super curious to hear what you think of it!
🚨 We're also hosting 2 live sessions TODAY:
•
10am EST | The Kilo Show for Non-Coders
We'll talk marketing automation, SEO, competitive analysis, and design
Just curious, when multiple subagents are running in parallel, how does the merge back to the parent agent work? Does it ever create conflicts when two agents touch overlapping parts of the codebase?
@boyuan_deng1 with the agent manager, each agent gets their own git worktree. That way they don’t conflict with one another. Then once they are done you can either open a PR directly from that worktree or have the agent merge it back into your local copy
@boyuan_deng1 There are two default subagents, general and explore. Explore tends to be utilized to find or answer a question about the codebase, and general can be used to make parallel edits but in a way that is directed by the parent agent to not cause conflicts (separate folders, modules, etc)
Great questions! Fun fact: parallel agents was one of the bets in v7 that took the longest to get right.
Try it here:
vscode:extension/kilocode.kilo-code
Looking forward to your thoughts
Report
How does Kilo handle context limits when you've got multiple subagents running on a large codebase? Does each subagent get its own context window or do they share?
@daniel_harris11 yes that’s the best part - each agent has their own context window. The parent agent passes only the context needed for the subagent to get its task done, and the subagent only passes the results back to the parent. All of the in between context doesn’t pollute one or the other
@daniel_harris11 Each subagent gets its own context, then provides a summarized report to the primary agent. You can view subagent sessions while they are running and the final report if you are interested.
@daniel_harris11 They each get their own when you're on separate worktrees (branching from the original agent), and can share when the work is merged back in :)
Been loving my experience with Kilo! Love the team and the speed at which things are shipped. This is just another great example of the speed at which this team can produce great work!
quick update: some users wanted more control back, and the team worked on multiple improvements, like better diffs before you approve and permission flow fixes.
curious about your pov - how do you like to work with AI coding agents? Human-in-the-Loop or agent-first?
A lot of people hit a breaking point when agents start generating more diffs than they can confidently review—how did you design the inline diff reviewer + line-comment-to-chat loop to reduce review fatigue, and what review metrics (time-to-approval, revert rate, “second pass” prompts) are you tracking to prove it works?
And yes, while you can bring your own keys or use local models, most users pay for inference via the Kilo Gateway, which lets you switch freely between models using one balance.
You just pay for token at provider costs - whatever Anthropic charges for Opus, and whatever OpenAI charges for GPT, that's what you pay!
Replies
Kilo Code
Hey Product Hunt 👋 - Job from the Kilo team.
Very excited for this launch. The new Kilo for VS Code is my daily driver, and this rebuilt version with agent manager is in my opinion the next step in agentic coding. I now let multiple agents run at the same time using agent manager, and it speeds up my workflow a lot. Super curious to hear what you think of it!
🚨 We're also hosting 2 live sessions TODAY:
•
10am EST | The Kilo Show for Non-Coders
We'll talk marketing automation, SEO, competitive analysis, and design
Register →
•
11am EST | The Kilo Show for Coders
We'll talk agent orchestration, codebase indexing, and IDE workflows
Register →
Mastra
@jobrietbergen yes, and we're hanging out here, on Twitter/X and LinkedIn all day - ask us anything ✌️
Does the agent manager have any limits on how many parallel agents you can run at once, or is it just constrained by your machine's resources?
Kilo Code
@barnaby_lloyd it is only constrained by the hardware you’re running on.
Mastra
@realolearycrew for example, what's your setup and how many agents do you run in parallel?
Kilo Code
@barnaby_lloyd It's constrained by machine resources
Mastra
Yes, and ICYMI you can read the docs for full details on parallel agents here: https://kilo.ai/docs/automate/agent-manager
KiloClaw
@barnaby_lloyd No limit, so it's a machine constraint if any!
Great update guys! Nice work!
Mastra
ty! what's your favorite new feature in this release?
Mastra
also curious what should @Kilo Code launch next? take the poll here in /p/kilocode
Kilo Code
@th_calafatidis thank you, happy you like it!
Mastra
@jobrietbergen anything we could share from the roadmap btw? any early previews on what's next for @Kilo Code?
KiloClaw
@th_calafatidis Thanks a lot!
Mastra
let's spread the word on X! repost this
Just curious, when multiple subagents are running in parallel, how does the merge back to the parent agent work? Does it ever create conflicts when two agents touch overlapping parts of the codebase?
Kilo Code
@boyuan_deng1 with the agent manager, each agent gets their own git worktree. That way they don’t conflict with one another. Then once they are done you can either open a PR directly from that worktree or have the agent merge it back into your local copy
Mastra
adding here the direct link to the docs for full context re: Agent Manager
https://kilo.ai/docs/automate/agent-manager
hope it helps - happy shipping!
Kilo Code
@boyuan_deng1 There are two default subagents, general and explore. Explore tends to be utilized to find or answer a question about the codebase, and general can be used to make parallel edits but in a way that is directed by the parent agent to not cause conflicts (separate folders, modules, etc)
KiloClaw
@boyuan_deng1 We use git worktrees to prevent conflict so that you can merge without issues!
Mastra
also curious how many agents do you run in parallel?
Mastra
Great questions! Fun fact: parallel agents was one of the bets in v7 that took the longest to get right.
Try it here:
Looking forward to your thoughts
How does Kilo handle context limits when you've got multiple subagents running on a large codebase? Does each subagent get its own context window or do they share?
Kilo Code
@daniel_harris11 yes that’s the best part - each agent has their own context window. The parent agent passes only the context needed for the subagent to get its task done, and the subagent only passes the results back to the parent. All of the in between context doesn’t pollute one or the other
Kilo Code
@daniel_harris11 Each subagent gets its own context, then provides a summarized report to the primary agent. You can view subagent sessions while they are running and the final report if you are interested.
KiloClaw
@daniel_harris11 They each get their own when you're on separate worktrees (branching from the original agent), and can share when the work is merged back in :)
Mastra
great question! also curious how many agents do you run in parallel?
Krater
Niiiice! Looks amazing and clean
Mastra
@maltepruser lfg! what new feature are you most excited about?
Mastra
@maltepruser also curious what should @Kilo Code launch next from your POV? take the poll here in /p/kilocode
Kilo Code
@maltepruser thank you!
Mastra
@maltepruser @jobrietbergen yes, and please help us spread the word on X - repost this
KiloClaw
@maltepruser Thank you!
Orbit
Been loving my experience with Kilo! Love the team and the speed at which things are shipped. This is just another great example of the speed at which this team can produce great work!
Mastra
spot on! this team keeps shipping: @Kilo Code for @VS Code and @JetBrains, and more with @KiloClaw...
Kilo Code
@erik_israni Kilo Speed!
Mastra
@jobrietbergen to quote The Leaderboard, @Product Hunt's daily newsletter:
Source: producthunt.com
KiloClaw
@erik_israni Thank you! Glad you're enjoying it :)
Used the old Kilo VS extension and have been using the new. Love the changes and it works smoothly. Excited to see many of the coming updates, too.
Mastra
Thank you, Josh!
Mastra
Follow-up question: What should we build/improve/fix from your perspective? Take the poll here in /p/kilocode
Kilo Code
@josh_slone1 thanks for joining us on the journey. What do you like most from the new version?
Mastra
@jobrietbergen following this 👀
Mastra
quick update: some users wanted more control back, and the team worked on multiple improvements, like better diffs before you approve and permission flow fixes.
curious about your pov - how do you like to work with AI coding agents? Human-in-the-Loop or agent-first?
KiloClaw
@josh_slone1 Thank you! Glad you're enjoying it.
Product Hunt
KiloClaw
@curiouskitty Good question!
We offer both local reviews in the extension and automated reviews on Github and GitLab PRs.
You can specify the strictness in both cases, so that you can filter down to the review points you actually care about.
Coldtea
How does the multi agent system comparison handle tokens ? does it run them all in the background simultaneously ? btw Congrats on the launch :)
KiloClaw
@farhan_nazir55 Thank you!
And yes, while you can bring your own keys or use local models, most users pay for inference via the Kilo Gateway, which lets you switch freely between models using one balance.
You just pay for token at provider costs - whatever Anthropic charges for Opus, and whatever OpenAI charges for GPT, that's what you pay!
Mastra
exactly - pay only for what you use. learn more about @Kilo Code pricing here: https://kilo.ai/pricing
Kilo Code
@farhan_nazir55 it runs them on separated git worktrees
Mastra
thanks for the support! help us spread the word on LinkedIn, repost this