Introducing Kilo for GitHub

Your newest AI agent has a GitHub handle

Agentic coding tools are improving rapidly, but most of them share the same friction: to use one, you have to leave the place you were working.

You spot something in a pull request, then you open your editor, pull the branch, paste the file into a chat window, and re-explain context the tool can’t see for itself. By the time the agent is caught up, you’ve spent more effort briefing it than the task was worth.

Code review and issue triage don’t live in your editor, though. They live in GitHub, in issue threads and PR diffs and review comments. So that’s where we put an agent.

@kilocode-bot is a coding agent that lives in GitHub. You can add it to any repository and mention it like a teammate. Tag it in an issue, a pull request, or a review comment, and it behaves like a colleague who has already read the entire thread.

Ask a question and it answers. Point it at a bug and it works out the cause. Hand it a fix and it opens a pull request - all without you leaving the tab you were already in. It reads the issue description, the diff, the review conversation, and the connected code on its own, so you brief it the way you would a person: with an @mention and a sentence.

A second set of eyes during review

The smallest and most frequent thing you’ll reach for it is a second opinion mid-review. You’re going through someone’s PR, you hit a block of logic that looks off or that you don’t recognize, and instead of context-switching to investigate, you ask right there:

@kilocode-bot is this true?

It reads your comment, the diff around it, and the code that the block depends on, then replies inline. Because it has the surrounding context, it can do more than answer yes or no: it’ll check whether the logic holds, suggest a cleaner approach if there is one, or explain an unfamiliar pattern a teammate wrote six months ago. It’s the reviewer who walked into the room already knowing the codebase, which is usually the part that makes a good second opinion expensive to get.

Bug Triage

Bug reports are easiest to act on once you know roughly what you’re looking at, and that first pass is often the slowest part of the job. When one lands, you can hand the agent the cold open:

@kilocode-bot what could be the cause of this issue?

It reads the report, searches the codebase for the code paths involved, and posts its analysis straight into the issue thread. You start digging with a theory instead of a blank page, and because the reasoning lands in the thread itself, whoever picks the issue up next sees it too. That makes it as useful for triage and assignment as it is for the person who eventually does the work.

From issue to pull request

When an issue is well described and you’d rather it just get handled, ask for the fix outright:

@kilocode-bot please fix

The agent reads the issue, analyzes the relevant code, writes the change on a new branch, commits it, and opens a pull request back to the repo. The PR shows up linked in the issue thread, and you review and merge it exactly like one from any other contributor.

How it works

Every mention spins up a Cloud Agent to do the work, which means it runs in the background rather than tying up your machine. You tag it and go back to what you were doing, and it comes back when it’s done.

The agent reads the full context it has access to, the issue description, the PR diff, the review thread, and your connected repository code, and then answers, analyzes, or opens a PR depending on what you asked for. It uses Kilo credits the same way every other Kilo interface does.

Getting the best results out of it

A few habits make the difference between a sharp response and a confused one. Be specific about what you want, since “please fix” rewards a well-described issue and stumbles on a vague one. If an issue is thin, ask the agent to diagnose before you ask it to implement. When a change spans multiple repositories, name them explicitly in your comment. And for quick, contextual questions, lean on review comments, where the agent has the tightest possible view of exactly the code you’re asking about.

Add it to your repos

You’ll need a Kilo account with available credits and your GitHub account connected through KiloConnect, which lives in the Integrations tab at app.kilo.ai. From there, install the Kilo Code Bot app on the repositories you want it active on. Once both are in place, @kilocode-bot is sitting in every issue and PR thread in those repos, waiting for its first mention.