A sandbox browser has none of your logins. Kortix in Chrome uses yours, so the controls matter:
agent tabs glow, Take over pauses it, Alt+Shift+K stops every agent, and it can be kept to tabs it opened.
Now on the Chrome Web Store. Experimental.
kortix.com/docs/connect/chro…
1
12
A quiet way to break production: deploy an older build over a newer one.
kortix apps deploy now records the git commit it ships and refuses a source older than the live version. --force if you mean it.
That check belongs in the tool, not in someone's memory.
kortix.com/changelog/2026-11…
3
Our agents used to refuse with "agent sessions cannot" and stop there.
Now a Kortix agent acts with its person's rights. When it can't do something, the refusal names the missing permission and who can grant it.
A refusal should tell you the next step.
kortix.com/changelog/2026-11…
1
2
78
Does an agent build Kortix now? About a third of it.
Nov 13–19 on our public dev branch: 234 commits. 75 authored by the Kortix Agent, 154 by four people, 5 by bots.
The agent runs the fix-and-review loop. People still made most of the commits.
kortix.com/?utm_source=x&utm…
1
38
Most agent setups assume your dev environment works.
Kortix's setup, new this week, proves it by running it, sizes the agents' computer to your stack, then resizes it from measured peaks.
An agent that can't run your tests can't fix your bugs.
kortix.com/changelog/2026-11…
3
65
Our reviewer agent says no a lot.
Nov 13–19 in our public repo, 62 of 147 agent reviews were "request changes".
This PR got blocked 5 times, each time with a repro or a failing test, before a 6th review approved it and it merged.
kortix.com/?utm_source=x&utm…
1
3
170
The PR, every review verdict included: github.com/kortix-ai/Company…
7
Not every agent needs a model.
Kortix now has program agents: an agent that runs your own code on every turn instead of a model, with a time limit and a payload input.
For the jobs where you want the same steps every time.
kortix.com/changelog/2026-11…
2
32
An agent's call can wait for your answer.
An Ask rule holds the action. Approve it in Review, a Slack thread or a Teams chat: it runs once, and Kortix resumes the agent with the result.
kortix.com/docs/guide#approv…
2
23
This week we posted how Kortix fixes Kortix overnight.
New this week: Loopmaxxing. Paste one prompt into Claude Code, Codex, Cursor, OpenCode or pi, and it adds a Software Factory to your repo: agents on your repo and tools, 24/7. Free to add.
kortix.com/changelog/2026-11…
3
90
Most AI at work is single-player: your chat, your prompts, your context.
A company's agents shouldn't be. In Kortix the whole team uses the same agents, skills, memory and connectors, shares a session and approves what the agent does.
kortix.com/multiplayer?utm_s…
1
31
A company's agent setup, as one file.
Agents for engineering triage, marketing and inbound sales, the tools each one gets, an approval rule for LinkedIn posts, and what wakes them: two webhooks and a weekly cron.
The example from Kortix's homepage:
kortix.com/framework?utm_sou…
1
41
Error at 02:52. Merged at 04:32.
One real fix from Nov 14:
02:52 a frontend error lands in Better Stack
03:50 an agent opens the PR (KRTX-4913)
04:19 a second agent approves it
04:32 merged, live on dev ~20 min later
kortix.com/?utm_source=x&utm…
1
1
36
The whole PR is public, test output included: github.com/kortix-ai/Company…
7
An agent that freezes until you answer it is a bad coworker.
In Kortix, agent questions and permission asks no longer block a session. After a short wait they park in Review, and your answer arrives as a new message.
kortix.com/changelog/2026-11…
24
Before IT lets agents near company systems, they ask the same six things:
who can use what, where an agent must stop and ask, what gets recorded, where it runs, how people sign in, and whether you're locked in.
Kortix's answers, on one screen:
kortix.com/security?utm_sour…
20
Kortix fixes Kortix. At any hour.
Our software factory is one loop:
a Better Stack error becomes a Linear issue,
an agent fixes it in its own sandbox and opens a PR,
a second agent reviews it, and it merges.
It runs on the same agents, triggers and sandboxes you get in Kortix: kortix.com/?utm_source=x&utm…
27



