Kilo Code, the open-source AI coding startup backed by GitLab cofounder Sid Sijbrandij, is launching a Slack integration that enables software program engineering groups to execute code adjustments, debug points, and push pull requests straight from their group chat — with out opening an IDE or switching functions.
The product, referred to as Kilo for Slack, arrives as the AI-assisted coding market heats up with multibillion-dollar acquisitions and funding rounds. However quite than constructing one other siloed coding assistant, Kilo is making a calculated wager: that the way forward for AI growth instruments lies not in locking engineers right into a single interface, however in embedding AI capabilities into the fragmented workflows the place selections truly occur.
“Engineering groups do not make selections in IDE sidebars. They make them in Slack,” Scott Breitenother, Kilo Code’s co-founder and CEO, stated in an interview with VentureBeat. “The Slackbot permits you to do all this — and extra — with out leaving Slack.”
The launch additionally marks a partnership with MiniMax, the Hong Kong-based AI firm that just lately accomplished a profitable initial public offering. MiniMax’s M2.1 model will function the default mannequin powering Kilo for Slack — a call the firm frames as an announcement about the closing hole between open-weight and proprietary frontier fashions.
How Kilo for Slack turns group conversations into pull requests with out leaving the chat
The combination operates on a easy premise: Slack threads usually comprise the context wanted to repair a bug or implement a function, however that context will get misplaced the second a developer switches to their code editor.
With Kilo for Slack, customers point out @Kilo in a Slack thread, and the bot reads the total dialog, accesses linked GitHub repositories, and both solutions questions on the codebase or creates a department and submits a pull request.
A typical interplay may appear like this: A product supervisor reviews a bug in a Slack channel. Engineers talk about potential causes. As a substitute of somebody copying the dialog into their IDE and re-explaining the drawback to an AI assistant, a developer merely sorts: “@Kilo based mostly on this thread, are you able to implement the repair for the null pointer exception in the Authentication service?”
The bot then spins up a cloud agent, reads the thread context, implements the repair, and pushes a pull request — all seen in Slack.
The corporate says the total course of eliminates the want to copy information between apps or soar between home windows — builders can set off advanced code adjustments with nothing greater than a single message in Slack.
Why Kilo says Cursor and Claude Code fall brief when builders want multi-repo context
Kilo’s launch explicitly positions the product towards two main AI coding instruments: Cursor, which raised $2.3 billion at a $29.3 billion valuation in November, and Claude Code, Anthropic’s agentic coding software.
Breitenother outlined particular limitations he sees in each merchandise’ Slack capabilities.
“The Cursor Slack integration is configured on a single-repository foundation per workspace or channel,” he stated. “Consequently, if a Slack thread references a number of repositories, customers want to manually swap or reconfigure the integration to pull in that further context.”
On Anthropic’s providing, he added: “Claude Code documentation for Slack exhibits how Claude could be added to a workspace and reply to mentions utilizing the surrounding dialog context. Nonetheless, it does not describe persistent, multi-turn thread state or task-level continuity throughout longer workflows. Every interplay is dealt with based mostly on the context included at the time of the immediate, quite than sustaining an evolving execution state over time.”
Kilo claims its integration works throughout a number of repositories concurrently, maintains conversational context throughout prolonged Slack threads, and allows handoffs between Slack, IDEs, cloud brokers, and the command-line interface.
Kilo picks a Chinese language AI firm’s mannequin as its default—and addresses enterprise safety issues head-on
Maybe the most provocative factor of the announcement is Kilo’s alternative of default mannequin. MiniMax is headquartered in Shanghai and just lately went public in Hong Kong — a lineage which will increase eyebrows amongst enterprise prospects cautious of sending proprietary code by way of Chinese language infrastructure.
Breitenother addressed the concern straight: “MiniMax’s latest Hong Kong IPO drew backing from main world institutional traders, together with Baillie Gifford, ADIA, GIC, Mirae Asset, Aspex, and EastSpring. This speaks to robust world confidence in fashions constructed for world customers.”
He emphasised that MiniMax fashions are hosted by main U.S.-compliant cloud suppliers. “MiniMax M2-series are world main open-source fashions, and are hosted by many U.S. compliant cloud suppliers akin to AWS Bedrock, Google Vertex and Microsoft AI Foundry,” he stated. “In actual fact, MiniMax fashions had been featured by Matt Garman, the AWS CEO, throughout this yr’s re:Invent keynote, displaying they’re prepared for enterprise use at scale.”
The corporate stresses that Kilo for Slack is basically model-agnostic. “Kilo would not drive prospects into any single mannequin,” Breitenother stated. “Enterprise prospects select which fashions they use, the place they’re hosted, and what matches their safety, compliance, and danger necessities. Kilo provides entry to greater than 500 fashions, so groups can at all times select the proper mannequin for the job.”
The choice to default to M2.1 displays Kilo’s broader thesis about the AI market. In accordance to the firm, the efficiency hole between open-weight and proprietary fashions has narrowed from 8 p.c to 1.7 p.c on a number of key benchmarks. Breitenother clarified that this determine “refers to convergence between open and closed fashions as measured by the Stanford AI Index utilizing main common benchmarks like HumanEval, MATH, and MMLU, not to any particular agentic coding analysis.”
In third-party evaluations, M2.1 has carried out competitively. “In LMArena, an open platform for community-driven AI benchmarking, M2.1 achieved a number-four rating, proper after OpenAI, Anthropic, and Google,” Breitenother famous. “What this exhibits is that M2.1 competes with frontier fashions in real-world coding workflows, as judged straight by builders.”
What occurs to your code if you @point out an AI bot in Slack
For engineering groups evaluating the software, a important query is what occurs to delicate code and conversations when routed by way of the integration.
Breitenother walked by way of the information stream: “When somebody mentions @Kilo in Slack, Kilo reads solely the content material of the Slack thread the place it is talked about, together with primary metadata wanted to perceive context. It does not have blanket entry to a workspace. Entry is ruled by Slack’s normal permission mannequin and the scopes the buyer approves throughout set up.”
For repository entry, he added: “If the request requires code context, Kilo accesses solely the GitHub repositories the buyer has explicitly linked. It does not index unrelated repos. Permissions mirror the entry stage granted by way of GitHub, and Kilo cannot see something the person or workspace hasn’t approved.”
The corporate states that information is not used to practice fashions and that output visibility follows present Slack and GitHub permissions.
A very thorny query for any AI system that may push code straight to repositories is safety. What prevents an AI-generated vulnerability from being merged into manufacturing?
“Nothing will get merged routinely,” Breitenother stated. “When the Kilo Slackbot opens a pull request from a Slack thread, it follows the identical guardrails groups already rely on at the moment. The PR goes by way of present assessment workflows and approval processes before something reaches manufacturing.”
He added that Kilo can routinely run its built-in code assessment function on AI-generated pull requests, “flagging potential points or safety issues before it ever reaches a developer for assessment.”
The open-source paradox: why Kilo believes making a gift of its code will not kill the enterprise
Kilo Code sits in an more and more widespread however nonetheless tough place: the open-source firm charging for hosted providers. The entire IDE extension is open-source underneath an Apache 2.0 license, however Kilo for Slack is a paid, hosted product.
The plain query: What stops a well-funded competitor — or perhaps a buyer — from forking the code and constructing their very own model?
“Forking the code is not what worries us, as a result of the code itself is not the hardest half,” Breitenother stated. “A competitor might fork the repository tomorrow. What they would not get is the infrastructure that safely executes agentic workflows throughout Slack, GitHub, IDEs, and cloud brokers. The expertise we have constructed working this at scale throughout many groups and repositories. The belief, integrations, and enterprise-ready controls prospects anticipate out of the field.”
He drew parallels to different profitable open-source corporations: “Open core drives adoption and belief, whereas the hosted product delivers comfort, reliability, and ongoing innovation. Clients aren’t paying for entry to code. They’re paying for a system that works day by day, securely, at scale.”
Inside the $29 billion “vibe coding” market that Kilo desires to disrupt
Kilo enters a market that has attracted extraordinary consideration and capital over the previous yr. The observe of utilizing giant language fashions to write and modify code — popularly referred to as “vibe coding,” a time period coined by OpenAI co-founder Andrej Karpathy in February 2025 — has turn out to be a central focus of enterprise AI funding.
Microsoft CEO Satya Nadella disclosed in April that AI-generated code now accounts for 30 percent of Microsoft’s codebase. Google acquired senior workers from AI coding startup Windsurf in a $2.4 billion transaction in July. Cursor’s November funding spherical valued the firm at $29.3 billion.
Kilo raised $8 million in seed funding in December 2025 from Breakers, Cota Capital, General Catalyst, Quiet Capital, and Tokyo Black. Sijbrandij, who stepped down as GitLab CEO in 2024 to focus on most cancers remedy however stays board chair, contributed early capital and stays concerned in day-to-day technique.
Requested about non-compete concerns given GitLab’s personal AI investments, Breitenother was temporary: “There are no non-compete points. Kilo is constructing a basically totally different method to AI coding.”
Notably, GitLab disclosed in a recent SEC filing that it paid Kilo $1,000 in trade for a right of first refusal for 10 enterprise days ought to the startup obtain an acquisition proposal before August 2026.
When requested to title an enterprise buyer utilizing the Slack integration in manufacturing, Breitenother declined: “That is not one thing we will disclose.”
How a 34-person startup plans to outmaneuver OpenAI and Anthropic in AI coding
Probably the most vital menace to Kilo’s place might come not from different startups however from the frontier AI labs themselves. OpenAI and Anthropic are each constructing deeper integrations for coding workflows, and each have vastly larger assets.
Breitenother argued that Kilo’s benefit lies in its structure, not its mannequin efficiency.
“We do not assume the long-term moat in AI coding is uncooked compute or who ships a Slack agent first,” he stated. “OpenAI and Anthropic are world-class mannequin corporations, and so they’ll proceed to construct spectacular capabilities. However Kilo is constructed round a distinct thesis: the laborious drawback is not producing code, it is integrating AI into actual engineering workflows throughout instruments, repos, and environments.”
He outlined three areas the place he believes Kilo can differentiate:
“Workflow depth: Kilo is designed to function throughout Slack, IDEs, cloud brokers, GitHub, and the CLI, with persistent context and execution. Even with OpenAI or Anthropic Slack-native brokers, these brokers are nonetheless basically model-centric. Kilo is workflow-centric.”
“Mannequin flexibility: We’re model-agnostic by design. Groups haven’t got to wager on one frontier mannequin or vendor roadmap. That is tough for corporations like OpenAI or Anthropic, whose incentives are naturally aligned with driving utilization towards their very own fashions first.”
“Platform neutrality: Kilo is not making an attempt to pull builders right into a closed ecosystem. It matches into the instruments groups already use.”
The way forward for AI-assisted software program growth might belong to whoever solves the integration drawback first
Kilo’s launch displays a maturing part in the AI coding market. The preliminary wave of instruments centered on proving that giant language fashions might generate helpful code. The present wave is about integration — becoming AI capabilities into the messy actuality of how software program truly will get constructed.
That actuality entails context fragmented throughout Slack threads, GitHub issues, IDE windows, and command-line classes. It entails groups that use totally different fashions for various duties and organizations with advanced compliance necessities round information residency and mannequin suppliers.
Kilo is betting that the winners on this market will not be the corporations with the finest fashions, however people who finest resolve the integration drawback — assembly builders in the instruments they already use quite than forcing them into new ones.
Kilo for Slack is accessible now for groups with Kilo Code accounts. Customers join their GitHub repositories by way of Kilo’s integrations dashboard, add the Slack integration, and might then point out @Kilo in any channel the place the bot has been added. Utilization-based pricing matches the charges of no matter mannequin the group selects.
Whether or not a 34-person startup can execute on that imaginative and prescient towards rivals with billions in capital stays an open query. But when Breitenother is proper that the laborious drawback in AI coding is not producing code however integrating into workflows, Kilo might have picked the proper combat. In spite of everything, the finest AI in the world would not matter a lot if builders have to go away the dialog to use it.
Disclaimer: This article is sourced from external platforms. OverBeta has not independently verified the information. Readers are advised to verify details before relying on them.
