Prosights.ai
← Products
AI Products · Kyberna

Features

Supporting Research

Kyberna 1.0.1 has 142 features and Kyberna Write 1.1.1 has 11, with 7 more coming soon. Each is listed here with what it does.

Kyberna Guard

32 features

Kyberna Guard More features are coming

Every model, every agent, under your rules.

Kyberna Guard checks what every model and agent does in Kyberna before it happens. A risky step, or one off your task, waits for your answer, and you can pause or stop everything at once. Kyberna's own models stay in the folders you chose, and mail and web pages reach the model marked as someone else's words, not yours. The browser never types a password, pays or accepts terms for you. Kyberna saves your files before a model or agent changes them, so you can put them back.

Privacy Protection

Frontier models can't see your personal data.

Kyberna redacts personal information before it reaches a model. Kyberna uses a built-in model and sophisticated techniques to find it. Kyberna redacts names, emails, phone numbers, street addresses, IP addresses, card numbers, IBAN bank account numbers, Social Security numbers, and individual taxpayer identification numbers. Kyberna returns the real values so you can see them. You keep your personal data where it belongs—with you.

Rogue Agent Protection

Kyberna watches an agent's every move and stops it in its tracks.

Kyberna blocks a harmful or off-task step by a model or agent before it runs. For example, reaching for your passwords, deleting outside the conversation's folders, tampering with Kyberna's safeguards, or performing a task you didn’t ask for. Kyberna not only tells you what the agent is trying to do but also provides a confidence score.

Kyberna holds Qwen3.6 35B's delete of a folder outside the conversation's folders, says it is 90% sure the model went off track, and asks: Don't Delete, Stop Qwen3.6 35B or Delete.

Protected Folders and Files Coming soon

Sacred cows protected at all costs.

Settings › Folder Access lists the folders no agent can delete, such as Documents, Desktop and Downloads. You add your own folders and files there, each at a level. No Delete: agents can change it but never delete or move it or anything in it, even inside a project. No Change (files) and Read Only: agents can read it but never change, delete or move it. The levels hold on every engine and for the commands you type with !.

Safe Rewrite

Where a safe form exists, the agent fixes the command before you see it.

Kyberna sends a dangerous command back to the model and asks it to rewrite it. Kyberna will let you know every time it asks the model to rewrite. Kyberna asks for at most 2 rewrites in one reply. A 3rd unsafe delete in the same reply waits for your answer. Kyberna will stop the agent from deleting your home folder, full disk, or the conversation's folder itself and ask for your permission. Kyberna also sends back a command that is cut off before its end. Examples are an unclosed $( or quote, or a line ending in | or &&. Kyberna then asks the model for the whole command. Claude Code 2.1.281 and Codex 0.154.0 do not tell the agent about such a command (read 2026-11-03).

Dangerous Delete Guard

Kyberna checks where a delete lands, not what the command says.

Kyberna refuses a dangerous delete before it runs, with no question to you. Protected folders include the full disk (“/”), your home folder, each conversation's folder, and Mac's temporary folder. Kyberna checks where a delete would land, even when the command hides the path in a variable.

Ask About This Request

Say no, and say why.

Kyberna may ask you to approve a step or hold one. Ask About This Request… lets you find out more about the model and agent's action before you answer. The request waits, and a panel opens beside it. You type a question, and the answer comes from the conversation without anything running. When you refuse the step, the words in the panel's field go to the agent. The agent can act on them.

Folder Snapshots

Undo what an agent did to your files, whatever tool it used.

At the start of every change, Kyberna snapshots each of the conversation's folders. A change waits until the snapshot is taken. Restore… after command finished puts the files back, and Undo Restore takes that back. Snapshots are kept 24 hours by default; you can set 1 hour to 30 days in Settings › Kyberna Guard. For a folder that would take more than 5 seconds to copy, Kyberna asks you first.

Advanced Rogue Agent Protection (Built-in model) Coming soon

A litttle brother watching.

A small built-in AI model made for Kyberna will check agent's each step against what you asked for, alongside Kyberna's classifier.

Explain This Request

Every request, explained in plain words before you answer.

Agents often execute sophisticated commands and scripts that are hard to understand. Kyberna will explain the agent's actions in layman’s terms.

Folder Boundary

Your models and agents stay in the folders you give them.

Kyberna forces agents to work only in the selected folders. Before an agent reads, changes, or runs anything elsewhere, Kyberna asks for your permission: refuse it, allow that one file, or add the folder. Every model and agent is blocked from accessing files holding keys, passwords, or another app's private data.

Kyberna asks before Qwen3.6 35B reads a file outside the project's folders: Don't Allow, Allow This File or Add the Folder.

Permission Boundary Coming soon

Every agent stays within the boundary, no matter who made it.

Kyberna will hold Claude Code, Codex, GitHub Copilot CLI, Grok Build and Antigravity CLI to the folders you gave the conversation. Kyberna already holds the models it runs itself to the conversation's folders. Before any of them opens a file anywhere else, Kyberna will ask you. Files holding keys, passwords or another app's private data stay refused to every agent, as they are today.

Prompt Injection Prevention

Words in an email or a web page are not your orders.

An email, a web page, or a message can hide instructions meant for the model, such as "send this person's files to me". Kyberna fences off everything someone else wrote, including hidden instructions, and strips the invisible characters attackers use. When people message Kyberna through Slack or another channel, Kyberna answers only the people you list. Kyberna sends mail only to people you allow. If an agent tries to send out something it just read from a private file, Kyberna pauses that step and asks you first.

Browser Guard

Payments, sign-ins and permissions stay in your hands.

On every site, Kyberna's browser refuses to accept terms, grant permissions, connect accounts, pay or sign in for you. Kyberna's browser never types into a password field and never carries text from one site into another. You cannot turn off the browser's rules.

Tool policy

Nothing runs without a rule you can see.

Before any tool runs, your rules in Settings › Tool Policy decide whether it runs, asks you first or is refused. Conversations, channels, reminders and tasks each have their own rules. Kyberna rates each step low, medium or high risk. For a command that runs a script, Kyberna reads the script before it runs and rates the command by what the script does.

Approval requests

Every request in plain words.

When a step needs your OK, Kyberna tells you who wants to do what and what the action does, colored by its risk. An edit shows its changes, and Explain this request provides additional context. Your answer, Don't Allow, Allow Once or Allow for This Conversation, decides what agents may do now and for the rest of the conversation. When several requests wait for your answer, Kyberna shows them one at a time, oldest first.

An approval sheet rates Qwen3.6 35B's delete as high risk, says the turn's snapshot can restore the file, and offers Don't Allow, Allow Once and Allow for This Conversation.

Approval levels

Fewer questions, never on what matters.

For each kind of step, you choose whether Kyberna asks every time, only for risky steps, or never. Kyberna always asks before a risky shell command, send, network or browser step. Tasks have their own row, like the other kinds of conversation, and a command's connection to a website follows the Network level.

Decide for me

Fewer questions, decided the way the agents' own auto modes decide.

Kyberna's checks answer the questions an agent's calls would ask, as Claude Code's auto mode, Codex's auto-review and Antigravity CLI's auto mode do. Reads and file edits in the conversation's folders run without a question. A call the checks find safe runs. A call they find unsafe is refused, and the agent is told why at once. Kyberna asks you about a call the checks find risky or cannot judge. Kyberna also asks about the first read outside the conversation's folders and any call your own rules cover. After 3 refusals in a row or 20 in a conversation, Kyberna asks you again until you allow a call. Under each reply you see what Kyberna allowed and refused. Allow This Call lets a refused call through when the agent makes it again. Kyberna's own checks answer for every engine, in place of each agent's own mode.

Remembered answers

Approve an action once per conversation, so your attention stays on the questions that matter.

When you choose Allow for This Conversation, Kyberna stops asking about the same kind of request until that conversation ends. Before you answer, Kyberna says what your answer covers, such as “git push” commands or edits in the conversation's folders. Kyberna forgets the answer when you close the conversation. Kyberna still asks before every email it sends, and sends email only to people on your send allowlist. A high-risk command you allowed for good runs with 1 orange warning line in the transcript. Codex (rust-v0.154.0) and OpenCode (v1.19.0) show nothing when an allowed command runs (compared 2026-11-01).

Package Sites

Installing a package just works, and nothing else gets out without your OK.

Kyberna lets commands in a conversation install from the official package sites (PyPI, npm and Homebrew) without asking. Before a command reaches any other website, Kyberna asks you. Kyberna checks the websites commands reach on Claude Code, Codex and every other command-line agent except Antigravity CLI. A website that only collects an agent's usage statistics is refused without asking. The agents other than Claude Code and Codex start only while Kyberna Gateway runs.

Site approvals

The model goes only where you let it.

A conversation asks before opening a website that neither you nor its search named: Don't Allow, Allow Once or Allow for This Conversation. Kyberna checks a redirect to another site again.

Safe websites

Trusted sites, no questions asked.

Sites you list open in a conversation's browser without asking, and a domain covers its subdomains. Browser Guard's rules still apply on every listed site.

Cookie banners

Kyberna never accepts cookies for you.

Kyberna never accepts cookies for you. A conversation using the browser presses the button that declines what a site does not need, or leaves every banner to you. Kyberna records each press.

MCP server checks

An MCP server cannot change behind your back.

When you turn on an imported MCP server, Kyberna reads its tools first and keeps it off while a tool's description reads like an instruction to the model. It also notes the server's command, files and version, and turns the server off when they change. When an agent other than Claude Code uses the server, Kyberna starts the server itself and asks you before each call, whatever the agent's own settings say.

Encrypted Record

Encrypted on your Mac, with a key that never leaves it.

Kyberna encrypts every conversation it records on your Mac, with a key Kyberna makes. Your Mac's Keychain holds the key for programs signed like Kyberna's own, and the key stays on your Mac. You set up nothing and type nothing. Nobody can read a copy of Kyberna's conversations taken off your Mac, such as a backup, a Time Machine volume or a synced folder. A program running under your login reads the record file and cannot open it without the key. While Kyberna is running, that program can still get conversation text through Kyberna's own command-line tool. Such a program cannot read Kyberna's memory. FileVault, the Mac's own disk encryption, protects a Mac that is lost or stolen.

Confirm it is you

Only you can take a whole conversation out.

Handing a conversation to a terminal, sending it to your other Mac, or copying over 50 rows asks for Touch ID or your login password. After you confirm, Kyberna doesn't ask again for 5 minutes for actions taken in the same place. Handing over and sending share a confirmation; copying has its own.

Loop detection

No endless loops on your time and dime!

Kyberna ends a turn when the model keeps repeating the same step with the same result. Kyberna also ends a turn when a local, server or API model's reply keeps repeating the same sentences.

Silent-turn watchdog

A stuck turn never runs forever.

Kyberna retries or resumes a turn that goes quiet and notes it in the transcript. Kyberna stops a turn that stays silent for 30 minutes. Time spent running a tool or waiting for your answer doesn't count against 30 minutes limit.

Pause and Kill a Session

Stop everything instantly. Suspicious activity ends on your command.

The Pause button freezes everything in a conversation, including its background agents and commands. The Resume button continues the conversation where it stopped. The Kill button, or ⌥⌘., ends everything the conversation is running after you confirm. Other conversations, reminders Kyberna is carrying out and jobs you scheduled keep running. A message you type during a pause goes out when you click Resume. After a Kill, messages still waiting to be sent go back to where you type.

Pause All Activity

One click holds everything.

Pause All Activity freezes every conversation Kyberna is running at once. You find Pause All Activity in the Kyberna menu and under Kyberna's icon in the menu bar, even with Kyberna's window closed. Reminders, scheduled jobs and messages from Slack or Signal wait until you choose Resume All Activity. A conversation you paused yourself stays paused after Resume All Activity. The menu-bar icon also lists each reminder, job or Slack request Kyberna held as suspicious, with Stop This Run and Let It Continue. Restarting Kyberna's background service also resumes everything.

Kill All Activity

The emergency stop for every agent Kyberna runs.

Kill All Activity ends everything Kyberna is running in every conversation at once. The ending covers background agents, background commands, scheduled jobs and reminders Kyberna is carrying out. You find Kill All Activity in the Kyberna menu and under Kyberna's icon in the menu bar. Kyberna asks you to confirm before ending anything. Unfinished work is lost. A scheduled job's history says when Kill All Activity stopped the job. Other Macs using your shared local model lose their current request. Kyberna's background service keeps running, so you can carry on right away. The Kyberna menu's Kill All Activity also works while Kyberna's background service is stopped.

Telemetry Off

Don't let your agents track you. Kyberna switches telemetry off wherever an agent allows.

In conversations recorded in Kyberna or kept Private, Kyberna turns off the usage statistics and other tracking data coding agents send their servers. Kyberna does not turn off GitHub Copilot's usage data.

Conversation record and privacy

7 features

Record modes

Encrypted by default. Private stays private.

Kyberna records conversation transcripts locally, encrypted. A conversation marked Private is not written to disk, and closing a Private conversation discards the transcript. After your first message, a conversation can move only to a more private choice, never back.

Private conversations

A conversation that leaves no trace.

A Private conversation writes nothing to disk and is discarded when its window closes. You cannot send a Private conversation to another Mac or hand it to a terminal.

Conversation retention

Keep conversations as long as you want, and no longer.

Keep conversations for sets how long Kyberna keeps conversations, Claude Code's transcripts included, and any conversation can set its own period from the record pill. A conversation answering an open reminder is never removed.

Recovery key

Your conversations survive a new Mac.

Show Recovery Key… shows the key that restores your conversations after you erase this Mac or move to a new one. Kyberna asks for your Mac's login before showing the key, lets you copy or print it, and never saves it to a file.

Move to a new Mac

Bring every conversation to your new Mac, and keep 1 key.

Make Move File… on your old Mac writes every conversation, Kyberna's and Claude Code's, into 1 encrypted file. Choose File… on the new Mac opens it with the old Mac's recovery key. Kyberna adds the old Mac's conversations, Kept meetings, projects, paused scheduled tasks and reminder records beside what is already on the new Mac. Private conversations are not moved.

Continue in a terminal

Start in Kyberna, finish in Terminal, and back.

Continue in Claude Code… hands a conversation to Claude Code in Terminal, and Take Back brings the turns you took there back into Kyberna. While it is in Terminal, the copy is plain text. A Private conversation cannot be handed over.

What each engine keeps

No guessing about where your words go.

A note in each conversation says what the model's maker keeps of your words, on this Mac and on its servers. Kyberna writes the note from the maker's published terms, and says so where the terms are silent. For a conversation recorded in Kyberna or kept Private, the note also names what Kyberna turns off for coding agents.

Reminders and tasks

14 features

Tagged reminders

Tag a reminder anywhere, and Kyberna gets the job done.

Add @Kyberna to a reminder in Apple's Reminders app on any of your devices. Kyberna on your Mac carries out the reminder at its due time, or right away if the reminder has no due date. Kyberna writes its answer into the reminder's notes. A ✓ at the start of the title and an alert on every device tell you the answer is ready. Kyberna never marks your reminder complete. You mark the reminder complete after reading the answer.

Reminder answers and follow-ups

A conversation inside your reminder.

Each answer in a reminder's notes is a block Kyberna keeps a copy of. A line you type under it gets a new answer, and an edit you make stays yours. Kyberna never rewrites your note.

On-device reminder routing

Simple reminders answered on your Mac, for free.

Apple's on-device model answers simple reminders for free and offline and hands anything needing tools, the web or exact facts to the model engine.

Excluded lists

You choose which lists Kyberna listens to.

Anyone who shares a list in Apple's Reminders app with you can tag Kyberna in that list. In Settings › Reminders, you can tell Kyberna to ignore chosen lists, such as a list shared with family. You can instead mark a list as shared with other people. Kyberna then treats the notes of a reminder in a marked list as someone else's words, not yours. An email to an address found only in those notes is saved as a draft, not sent.

Reminder runs

See exactly what Kyberna did for each reminder.

Settings › Reminders and the sidebar list each reminder's past runs, with Re-run, Open in Reminders and the conversation that answered it.

Reminder outcome on Slack

Hear back wherever you are.

Each finished reminder can also go as a direct message to your Slack channel's operator.

Reminders on two Macs

2 Macs, 1 answer.

When 2 Macs watch the same lists, only 1 of them answers each reminder.

Reminders test mode

Try reminders without the buzz.

Test mode (no alarms) writes the outcome and the ✓ but sets no alarm, so trying reminders out does not alert your devices.

Scheduled tasks

Put recurring work on a schedule.

A task is a prompt Kyberna runs once, every so many minutes or hours, daily, on weekdays or weekly, each run in its own conversation. You choose what it runs on from the same list as a conversation's engine pill, or leave it on the default model. A scheduled task runs while you are logged in and the Mac is awake.

Task catch-up

Asleep at the scheduled time? A missed task runs when your Mac wakes.

If the Mac slept through a task's time, the most recent missed run goes once on waking, within 7 days. You can choose to skip it instead. At most 3 runs go at once.

Unattended task rules

Unattended, but never unsupervised.

A task runs under its own tool rules: sending mail, shell commands and file edits wait for your approval, and it never uses the browser. A run nobody answers is given up at the next scheduled time, except a run Rogue Agent Protection paused, which waits for you.

Task history and notifications

Every run, and what it found.

A task's page lists its last 100 runs with their outcomes and conversations. Kyberna Gateway notifies you when a run finishes, fails, waits for you or catches up, even with the Kyberna window closed.

Paused background runs

Paused where it runs, answered from wherever you are.

When Rogue Agent Protection pauses a reminder, task or channel run, a notification offers Stop This Run and Let It Continue. The notification comes even with the Kyberna window closed, and the menu-bar wheel lists the paused runs. You can also answer in the reminder itself with a code, in lists you allow, and a channel's operator answers in a direct message. A paused run waits for you with no time limit; restarting Kyberna Gateway stops it.

Reminder file access

Reminders read the files you allow, and ask on your phone for the rest.

A reminder reads files in its own folder and in Folders reminders may read, and asks before reading anywhere else. You answer on the Mac (once, for that reminder, or always) or, in lists you name, by replying in the reminder with a code. A folder of private data, or your whole home folder, can be allowed only on the Mac and only once at a time.

Local models

15 features

Local models

Enjoy open weight models. No account, no cloud, no one collecting your data.

You can run local models within Kyberna on your Mac, and your information stays local. Kyberna lets you download LLM models. Kyberna also finds models already downloaded in its own folder, in the LM Studio and Ollama folders, and in any folder you choose. Kyberna lists certified models and tells you whether they will fit your Mac's memory.

Model Catalog

Find a model that fits your Mac.

The Model Catalog (⇧⌘L) searches Hugging Face, shows whether each model fits your Mac's memory, and downloads it, checking each file against its published checksum. Kyberna does not screen what a model file contains.

Hugging Face token

Your Hugging Face access, in Kyberna.

A stored Hugging Face token raises the rate limit and opens the gated repositories your account has access to.

Loaded Models

See what your Mac's memory is doing.

The Loaded Models pane shows your Mac's memory and every model loaded now, in Kyberna, LM Studio and Ollama. The Unload button in Loaded Models waits while a conversation is answering.

Memory check

No more guessing whether a model will fit.

Kyberna checks your Mac's available memory before a local model loads. A window warns you if the model would leave memory tight or would not fit. The window offers to unload idle Kyberna models or to quit LM Studio and Ollama to make room. Kyberna closes nothing until you click. Your message waits until the model is cleared to load.

Context and thinking

Configure a local model for speed and depth.

For local models, you choose how much history it can hold, from 8K up to the model's own length. You also set whether to use thinking mode. Kyberna maintains persistent context so the context is not lost when app restart. Thinking returns to the model's default after the restarte. For the models other apps use through Kyberna, Settings › Models sets each model's context size, kept when Kyberna quits.

Certified models

Certified to do the work, not just to talk.

Kyberna marks a local model "Certified for tools" after the model passes Kyberna's 6-prompt tool test. The 6 prompts ask the model to summarize email, list changed notes, read code files and handle an email that a rule refuses. A certified model calls the right tool every time and answers from what it read. Qwen3.6-35B-A3B and Qwen3.5-35B-A3B pass the test on a Mac with 32 GB of memory. A model without the label still chats with you in a conversation.

Core AI models

Apple's on-device runtime, for open models.

Models already converted to Apple's Core AI format run on Apple's own on-device runtime; Kyberna doesn't convert models itself.

Local model speed

Follow-ups answer faster.

A local model's result line shows how fast it answered, in tokens per second. The result line also shows how much of the conversation the model reused from the turn before.

Commands run by local and API models

Your local model does real work, inside a fence you set.

Kyberna lets a model on your Mac run Terminal commands in the folders you gave the conversation. Models in LM Studio or Ollama and cloud models reached through the Anthropic or Gemini API can run commands too. Kyberna asks you before any risky or unknown command runs. Each command runs in a sandbox that can write only in the conversation's folders. A command that tries to read or write anywhere else fails. Commands reach the internet only through Kyberna, which asks you about any site other than package downloads. A command cannot use Kyberna's access to your Mail, Notes, Calendar, Contacts or Reminders. The Pause, Resume and Kill buttons reach every command, including commands running in the background.

Web pages Kyberna's own engines read

Your local model can surf the web.

A local model or the cloud model with API can read a web page whose address starts with https://. By default, Kyberna asks before each page and names the website. The model gets the page's words marked as text from that website. A long page comes in parts the model reads one after another.

Background tasks list

Every command your model left running, in one place, yours to pause or stop.

When your model starts commands in the background, a button in the conversation's header counts them. The button opens a list where you read, pause, resume or stop each background command (⌥⌘P). Claude Code and the other command-line agents run their own commands, which the list does not show.

Agents a conversation starts

Your model can hand work to agents, and you see and steer every one.

When a model or agent starts helper agents, each carrying out one task, a list above the message field shows them. When they end, the conversation replies with their results with no message from you, on every engine. The conversation's model is told of each pause, resume, stop and restart. An agent silent for 10 minutes is restarted, up to 5 attempts; on a CLI the limit is 16 minutes. An agent that ends during a running turn leaves a row after that turn. The list shows each helper agent's state, name, time and what it is doing, with Stop and Review…. You can open any agent's conversation, and its approval requests and holds come up on the conversation that started it, naming the agent.

Messages between conversations

Your conversations can talk to each other, and you decide what each one takes.

You can read and send messages to other converrsations, except Private conversations. You can send a message to a convesation directly (@conversation) without using model and wasting tokens.

Serve Kyberna’s local models to other Macs and devices

Serve the models Kyberna runs to your other Macs and devices on the network, each with its own key.

Share Kyberna's Local Models lets other apps on this Mac use Kyberna's local models, each app with its own key you can remove. Serve on the Local Network, off by default, lets your other Macs and devices use them too. Kyberna on another Mac connects encrypted. Other apps use an unencrypted address, and the pane says so. When another request is using the model, a conversation waits its turn and says so.

Engines and models

10 features

Engine picker

Select right model for the job.

The engine pill picks the model and account for the conversation, even mid-conversation. Long model lists show the newest model of each family first. Every other model waits under More Models.

Default model and effort

Start every conversation on the model you prefer.

Settings › Models › Default model sets the engine, model and effort new conversations start on. The effort pill appears only for engines that take an effort level.

Existing subscription

Use the AI plan you already have.

You can use your existing Claude, OpenAI, Gemini, Grok, and GitHub subscriptions.

APIs

Pay per token when that suits you better.

You can use your existing Claude, OpenAI, Gemini APIs.

Cloud provider

Claude through cloud providers.

Use Claude through Amazon Bedrock, Google Vertex AI or Microsoft Foundry.

Model servers

Bring the model server you already run.

Models you run in LM Studio, Ollama, a service such as OpenAI or another server answer conversations. Kyberna finds LM Studio and Ollama on this Mac, and you add any other server by its address. A server with a self-signed certificate works only once this Mac trusts it.

Kyberna tools switch

Give a conversation your apps, or keep them out.

The last row of the engine pill turns Kyberna's tools (Mail, Notes, Calendar, Contacts, Reminders) on or off for a conversation. Kyberna's tools are on by default for Claude Code, the API, agents and certified local models, and off for other local models and servers.

Project files and instructions

Every model knows your project.

A model on this Mac, a model server or the Anthropic API reads the files in the conversation's folders, images and PDFs among them. A PDF comes as text, up to 20 pages at a time. Text recognition on this Mac reads scanned pages. An image comes as the picture with the words recognized in it. A model that does not see pictures gets the words alone. The model also follows your project's AGENTS.md or CLAUDE.md.

Automatic compaction

Long conversations keep going.

When a conversation on the API, a local model or a server fills most of its context, Kyberna summarizes the middle. Kyberna keeps the start and the recent part word for word. The transcript says when it did.

MCP servers

Bring the MCP servers you already use.

Import a Claude Desktop, Cursor or Cline configuration and its MCP servers come with it, each off until you turn it on. The servers' tools reach conversations on Claude Code, Antigravity CLI and the other external agents, and every call asks you. In a reminder or a scheduled task, each call waits for you, and a channel never gets them.

Other engines

3 features

Kyberna's Node.js

No developer tools to install first.

The first time an agent needs Node.js, Kyberna downloads and checks its own copy (about 50 MB), so you have nothing to install. Kyberna asks before removing it.

Engine downloads

The version Kyberna tested, and no silent updates.

Kyberna downloads each command-line agent at the version Kyberna tested. Kyberna checks the file before it runs. Setup and Settings show each engine's size and download progress.

Agent sign-in

Sign in with the vendor, not with Kyberna.

Sign In… opens the agent maker's own login in Terminal, so Kyberna never sees your password or token.

Conversations

21 features

Conversations

Work with any model on the files in your folder.

A conversation is 1 exchange with a model in 1 folder on 1 engine. You start it with ⌘N in the working folder or in a folder you pick each time.

Recently Deleted

Deleted by mistake? Bring it back.

Delete moves a conversation, a scheduled task or a meeting out of the sidebar without asking. View › Recently Deleted lists them for 30 days with the days left. Recover puts each back at once. Delete Immediately deletes the selected items for good after a confirmation. A deleted task does not run while it waits. After 30 days Kyberna deletes each for good the next time it is open. A task goes with its run list and its runs' conversations. A meeting goes with its notes' conversation. A Private conversation keeps nothing, so its delete asks first and is for good. Deleted reminders stay in Reminders' own Recently Deleted.

Start page

Start working the moment the window opens.

With no conversation selected, the main window greets you by name. The start page starts a conversation from what you type or say, on the model, folder and record choice its 3 pills set. The start page also offers your latest conversation and up to 4 places to start: your working folder, Mail, Reminders and the browser.

Sidebar filter and search

Every conversation, from any engine, in 1 list.

The sidebar groups conversations by task, reminder, project and folder, filters them by engine and searches their titles and folders.

Accounts and keys

Always know which account a conversation bills to.

The foot of the sidebar names the account the selected conversation bills to. The foot of the sidebar also lists every sign-in and key Kyberna can use, without ever reading a key or token.

Composer

Write, paste and attach the way you expect.

Return sends, Option-Return or Shift-Return adds a line, and Up brings back earlier messages. The paperclip attaches up to 20 images, PDFs and text files. The engine gets what Kyberna read from each file, never its folder. A long paste goes as quoted material unless you choose to send it as written.

Message queue

Keep talking while the model works.

A message you send while the model works waits as a bubble. Kyberna sends the waiting message when the running steps finish. Send now delivers it at once.

Model and skill commands

Switch models without leaving the keyboard.

Typing /model lists or switches the model without sending text to it. Typing /pace fast or /pace natural sets how the browser is driven. With the Anthropic API, a local model or a model server, /compact shortens the conversation now. On Claude Code, / and a skill's name runs that skill.

Status pill

See your usage before you run out.

The pill at the foot of a conversation shows how full the context is and your Claude usage, with the conversation's state. The status pill's rings turn orange at 70% and red at 90%.

Readable transcript

See what the model did, in plain words.

The transcript shows each step the model took as its own row and ends each turn with a result line. The transcript folds each browsing run into plain-language lines. Jump to Latest and View › Scroll to Top and Scroll to Bottom reach either end of a long conversation in 1 action.

Tool call rows

Every step the model takes, in its own words.

Each step in the transcript is its own row, titled with the step's own description or with the tool and what it works on. The row opens to what it ran and its result.

Command block

The exact command, ready to copy.

An open step shows its exact command, with shell coloring, or the file it works on, with a Copy button.

Tool call state and time

See what is running, and for how long.

Each step's row shows a spinner and a running time while it runs. When it finishes, it shows a check and how long it took. A failed step shows a red cross, and a step that needs your answer reads "Waiting for you".

Turn status line

Watch the turn's progress as it happens.

While a turn runs, a line under the transcript shows the time so far and the tokens written. The status line also shows the steps running and what the model is doing now.

Cost and provenance

Know which model answered and what it cost.

Each turn's result line gives its outcome, its steps and, when billed per token, the cost the engine reported. The result line's tooltip names the engine, provider and model. Only Claude Code reports a dollar figure.

Find in transcript

Find anything you discussed.

⌘F finds text in the open conversation, with Return and Shift-Return for the next and previous match.

Copy Transcript

Take a conversation anywhere as Markdown.

Edit › Copy Transcript copies the conversation as Markdown, or only the rows you choose with Start Range Here. Copying more than 50 rows asks you to confirm it is you.

Import Chat

Bring your claude.ai chats with you.

File › Import Chat… reads a claude.ai data export, and Edit › Paste as New Conversation turns a copied transcript into a conversation.

Ask Kyberna about Selection

Ask about any text, in any app.

Edit › Ask Kyberna about Selection (⌥⌘K) asks about selected text in Kyberna or Kyberna Write. Other apps reach the same panel through Services › Ask Kyberna.

Run a command yourself

Run a command beside your agent, and choose what it reads.

A line starting with ! runs on your Mac as you, in the conversation's folder, with no model turn. A row shows the output. The output goes with your next message unless you click Leave Out. Pause, Kill and Kill All reach the command. Running one never widens what the agent may reach. Compared 2026-11-04 with Claude Code as documented (pin 2.1.281) and Codex rust-v0.154.0, Kyberna spends no model turn by default. Claude Code answers each command unless turned off. The output reaches the agent marked as untrusted. A protected place's output stays with you. A Private conversation's output is held in memory alone.

Token Optimization Coming soon

Your model reads only what your task needs.

Token Optimization will let you take chosen prompts and answers, attached files and file reads out of what the model reads. Kyberna will keep each left-out item in the conversation as your record, with Put Back beside the item. You will be able to send 1 message to a model that reads none of the earlier conversation. Each attached file and each file read will show about how many tokens the model reads for it. Conversations from scheduled tasks and reminders will offer Token Optimization too. Token Optimization will work with Claude Code, Codex, GitHub Copilot CLI, Grok Build, Antigravity CLI and Kyberna's own models.

Projects and folders

4 features

Projects

Keep related work together.

A project groups conversations and the folders they work in. You rename it in place, archive it or delete it, and no folder or transcript is touched.

Folder switching

Move a conversation to the folder it needs.

The folder pill moves a conversation to another folder, except while a turn runs. Claude Code continues the same session, and other engines continue from a summary.

Folder trust

Trust a folder once, before any model opens it.

Before a conversation opens in a folder you haven't trusted yet, Kyberna asks whether you trust its files. Trust all folders trusts a project's folders at once.

Missing-folder recovery

Moved a folder? Pick up where you left off.

When a conversation's folder was moved or deleted, the conversation stays readable and asks you for the new location or a scratch folder. The conversation then continues there with its earlier turns.

Mail, Notes, Calendar and Contacts

8 features

Mail tools

Read, write, summarize your emails without worring about personal data.

A model can search and read your mail, save attachments into the conversation's folder and draft replies. Kyberna's mail tools never move or delete mail, and an email's text reaches the model marked as someone else's words.

Send allowlist

Mail goes only to people you allow.

Kyberna sends mail only to addresses you approved or in reply to your own message, and saves anything else as a draft, saying why. In a conversation, Kyberna asks before every email. The sheet names where an address came from when it didn't come from you. A message to more than 20 people becomes a draft, and a file holding passwords, keys or browser data is never attached.

Kyberna asks before sending an email to an address not on the allowed list, showing the recipient, subject and message.

Send checks

A message leaves with only what you meant to send.

Before a message leaves your Mac, Kyberna checks what goes with it. An attachment must come from the conversation's folders and hold no passwords, keys or programs. A web address or an access key in the text must come from you. A file that fails stops the message, and anything else that fails asks you or becomes a draft.

Notes tools

Notes that write themselves, and only add.

A model can search and read your notes, create a note and add a dated section to one. Kyberna's notes tools never delete or rewrite a note, and leave protected, shared or just-changed notes alone.

Calendar tools

Plan around your real calendar.

Once you allow Calendar, conversations and reminders see your events and can create or move one; in a conversation, creating or moving asks first.

Contacts lookup

The right address, never a guess.

After you allow Contacts, a model turns a name into addresses and phone numbers, read only. Asked about "me", it reads your own card (your My Card in Contacts). When nobody matches, Kyberna tells the model so instead of letting it guess.

Reminders tools

Your to-dos, in the conversation.

With Reminders access, a conversation lists, creates, updates and completes your reminders.

Writing guides

Updates and memos that lead with what matters.

Ask for a weekly update, a meeting summary, an email, a memo, a report or a proposal. The model follows Kyberna's guide for that kind. The writing puts the main point first, keeps the kind's shape and length, names its sources, and says plainly what is done and not done. Reminders and scheduled tasks use the guides too; channels don't.

Browser

7 features

Kyberna Browser

A browser built for working with AI.

Kyberna's own browser window takes Safari's keyboard shortcuts and keeps your sign-ins, but its address field takes addresses and does not search. Kyberna Browser identifies to websites as the Safari on your Mac, so Slack and other sites that refuse unknown browsers open.

Browser downloads

Downloads, the way Safari does them.

A link to a file saves it to your Downloads folder, numbered when the name is taken. The downloads list never opens on its own.

Private browser window

Browse privately, even with AI.

File › New Private Window (⇧⌘N) opens a window that keeps no cookies, history or site data once it closes. A conversation uses it only when you ask for private browsing.

Browser use in conversations

Ask, and Kyberna browses for you.

Ask a conversation to browse and it reads pages, clicks, types and moves between tabs out of sight. A picture of the page shows in its header. When a page needs you, it waits. You can take the page over and hand it back. Channels never browse.

Scroll to load

Reads the whole list, not just the first screen.

A conversation scrolls a page and presses Show more to read a whole list. Kyberna stops scrolling after 60 scrolls or 60 seconds unless the model asks for more.

Browser pace

Browse at your pace.

The Web pill switches a conversation's browsing between Fast and Natural Pace, which moves and types like a person. Natural Pace doesn't change what a site's terms allow.

Guided Slack setup

Set up Slack with a guide at your side.

The guided browser walks you through Slack's website 1 step at a time and stops on every step that is yours to do. The guided browser never reads a token or password and never accepts terms.

Channels

8 features

Channels

Your Mac's AI, from your phone.

Message Kyberna from Slack or Signal and get an answer back, with nothing on the Mac opened to the internet. A channel starts off with nobody allowed to write.

Slack channel

Kyberna in your Slack.

Make a Slack app from Kyberna's manifest and Kyberna answers in your workspace; its 2 tokens stay on this Mac.

Signal channel

Text your Mac on Signal.

Kyberna reaches Signal through a number of its own or as a linked device, using a signal-cli it downloads and checks (45 MB). A linked device cannot answer the account's own owner.

Channel access and operator

Only the people you choose.

Only the people you list may write to a channel, and Kyberna ignores messages from anyone else. Only 1 operator may approve tool calls or change the model.

Channel modes

A channel does only what its mode allows.

Read-only, Assist and Full set what a channel's conversations may do with the tools a channel is given. In no mode does a channel reach your mail, notes, calendar, contacts, reminders, the browser or your MCP servers. On a channel, an agent's own shell and editors are denied unless Settings › Tool Policy allows them. A channel on Codex, Copilot CLI or Grok Build reads its folder only through Kyberna's file tools. Kyberna's file tools never open a protected file or anything outside the folder. With Files set to Deny for Channels in Settings › Tool Policy, none of them reads a file. Claude Code and local models read no file on a channel. A call that waits for your OK is sent to the operator with a 4-digit code. The operator answers with allow or deny and that code, which works once. A call that reaches private files is allowed only on the Mac. With no operator set, the call is answered on the Mac, where the conversation and a notification say so. The sender is told only that the request waits.

Channel commands

Run Kyberna from a chat window.

The channel's operator switches the model, mode, effort and reasoning with /model, /mode, /effort and /reasoning (also /thinking). /help lists the commands. A waiting step is answered with allow or deny and its code; a bare 1, 2, yes or no lists what waits.

Offline catch-up

Old messages never trigger surprise actions.

When the Mac comes back, Kyberna records messages older than 15 minutes rather than acting on them. Kyberna tells each sender once to resend what they still want done.

Slack step-by-step guide

Message Kyberna from Slack in about 20 minutes.

1 page of the guide walks through installing Slack, creating a free workspace and the Slack app from Kyberna's manifest, and connecting it. The Slack setup takes about 20 minutes and a personal email address.

Voice

5 features

Dictation

Talk instead of type, privately.

Press the microphone and what you say is written into your message on this Mac, with nothing sent anywhere. Say "Kyberna, send" or "Kyberna, cancel" to finish. Kyberna asks for the microphone only the first time.

Read answers aloud

Listen to your answers.

Kyberna speaks a turn's finished answer in the Mac's own voice, only for the conversation on screen.

Voice mode

Hands-free, start to finish.

Voice mode listens until you end it and sends each message after a pause or on "Kyberna, send". The opt-in "Hey Kyberna" (or "Hi Kyberna") keeps the microphone on whenever Kyberna is open.

Meeting notes

Your meetings, written down on your Mac.

Press Start and Kyberna writes down your Zoom, Microsoft Teams or Google Meet call on this Mac. Kyberna then writes a summary, decisions and action items with the model you chose. You choose it in Settings for every meeting, or with Change… for one meeting. When that model has no quota left, Kyberna moves to the next model on your list. Kyberna labels your lines You and names the others from the calendar event; a meeting is Kept or Private, and the sound is never saved.

Meeting action items

From what was said to what gets done.

1 review sheet shows each action item as the reminder, calendar event, note or email draft it becomes, with its text, owner and date editable. The ticked items are added exactly as shown, with Undo, and an email is only ever a draft.

Session sync

2 features

Session sync

Your conversations on both Macs, encrypted.

Your conversations are copied, encrypted, into a folder your sync service carries to your other Mac. No file name shows a path or title, and a conversation already on the other Mac is never overwritten.

Send to other Mac

Hand a conversation to your other Mac.

Send to other Mac writes 1 conversation to the sync folder after you confirm it is you, sending only what is new each time.

Kyberna Write

12 features

Kyberna Write

Markdown, beautifully simple.

Kyberna Write is a Markdown editor that shows a document rendered, as source or both. Kyberna Write remembers your choice for each file and saves as you go. Kyberna opens its Markdown files in it.

Markdown window in Kyberna

Edit what the model wrote, right in Kyberna.

Kyberna opens Markdown files and linked files from a conversation in its own editor window, the same editor as Kyberna Write's.

Formatting bar

Format without memorizing Markdown.

In Source and Both, a bar formats headings, bold, italic, strikethrough, code, links and lists.

PDF and HTML export

Share it as PDF or a web page.

A document exports as PDF, prints, or exports as 1 self-contained HTML page whose images keep the paths the document names.

Linked documents

Follow your notes from page to page.

A link to another Markdown file opens it in its own window, and web links open in your default browser.

Ask Kyberna in Write

Ask your document anything.

A panel beside the document asks Kyberna about the selection or the whole document, under your tool rules; it needs Kyberna Gateway running.

Writing actions

Rewrite, translate or outline in 1 click.

Rewrite, Summarize, Translate, Continue and Outline offer their result as Replace Selection or Insert Below, and nothing changes until you click one.

Look up in this folder

Every claim, checked against your sources.

Kyberna finds the passages in the document's folder that support a selection, with file, lines and quote. Kyberna says which claims nothing supports, and it reads only a folder you trusted.

Writing Tools

Apple Intelligence, in your Markdown.

Apple's Writing Tools proofread, rewrite or summarize selected Markdown as plain text, on a Mac with Apple Intelligence turned on.

Outline

See the shape of your document.

The sidebar shows the document's headings and scrolls to the one you click.

Word count

Know how long it reads.

The line under the file name shows the words in the document and a reading time at 200 words per minute, not counting Markdown markers.

Send to Kyberna

From draft to conversation and back.

Send to Kyberna opens a conversation in a saved document's folder about that document, and Kyberna opens a conversation's file in Kyberna Write.

Setup and help

12 features

Setup sheet

Set up in 1 sheet, with nothing to configure by hand.

On first launch, setup asks 1 thing per page. The pages cover your name, the agents and local models to download, and the default model. The last pages cover the working folder, how conversations are kept, and the recovery key. Set Up Later keeps your choices, and Settings › General › Run setup again… opens it again.

Kyberna Gateway

Always on. Kyberna keeps working even when app is closed.

Kyberna Gateway runs in the background under your account, so conversations keep going after the window closes and reminders, channels and scheduled tasks run. Without it, a conversation ends when its window closes.

Kyberna Guide

User guide.

Help › Kyberna Guide (⌘?) opens the guide with a searchable contents list, and each Settings pane's ? opens the section about that pane.

The research behind Kyberna Coming soon

Behavioral science and AI research that powered the Kyberna's design.

Help › Kyberna Guide lists the published research behind each feature area and counts the other sources. Each published paper was read in its own text. The features page shows the same sources, 1 closed list per area.

Liquid Glass appearance

Designed for macOS 27.

The sidebar, header pills, composer, sheets and menus draw in macOS 27's Liquid Glass over a plain transcript. Settings › Appearance changes the transcript, not the glass.

Transcript font and size

Read your conversations in the type you like.

Settings › Appearance sets the transcript's font and a size from 10 to 24 points. Settings › Appearance also sets the size of the message field's microphone, send and Stop symbols.

Transcript look

Your messages in your color.

Settings › Appearance › Look offers Accent, your messages in your accent color in 13-point text, and Classic, the earlier gray look in 15-point text. Every install opens in Accent, and a look or size you choose is kept.

Working folder

Your files land where you can find them.

New conversations run in your working folder, ~/Kyberna unless you choose another in Settings › General, or in a folder you pick each time.

Developer mode

Every detail, when you want it.

Developer mode in Settings › Developer shows each conversation's folder and origin, and more detail in the status text. Developer mode also shows the engine, provider and model on each result line. With it on, Self build mode applies to projects marked for self build while agents build Kyberna. In those projects, commands and file changes inside the project's folders run without asking. Commands in a self build project still run inside Kyberna's sandbox. Every approval that is left shows with no offer to stop. The network, sending and the browser still ask.

Known limitations and defaults

What Kyberna does not do yet, stated up front.

1 page of the guide gathers every limitation and default behavior the guide documents. Each comes in the words of the section it comes from, with a link back to it.

Auto Updates Coming soon

New versions arrive by themselves.

Kyberna and Kyberna Write look for a new version once a day and download it in the background. The new version installs the next time the app quits and opens. Settings has 2 switches for this and Check Now, and each app's menu has Check for Updates… after Settings…. The first launch of a new version restarts Kyberna Gateway, interrupting what runs at that moment.

Settings search Coming soon

Type a setting's name and go straight to it.

Type in the Search field at the top of the Settings sidebar to list each matching setting. Each setting shows the name of its pane beneath it. A word from what a setting does finds it too. Choosing one opens that pane scrolled to the setting. Your search stays in the field. Clearing it lists every pane again. A setting is listed only while its pane shows it.

← Back to Products