Questions about Kryex, answered.

Kryex is a governed AI coding agent for enterprises. It has three parts: Kryex Code, the desktop coding agent developers use; Kryex Server, which authorizes every governed action and writes the audit log; and Kryex Admin, the console where your organization decides who can do what, on which system, with which model.

Kryex Code is a desktop AI coding agent for Windows and macOS. It plans the work, edits code, runs commands, drives a browser, delegates to subagents and remembers your project, on the model you choose. Anything that touches a governed system — a database, an MCP tool, the browser or web research — is checked on the server before it runs.

Kryex Admin is the control plane for coding agents. Admins approve access requests from one inbox, grant people READ, WRITE or ADMIN on each database and tool, set the command policy for every developer's agent, choose which models people can use and how much, and read a hash-chained audit log of every decision.

A governed coding agent can't grant itself access. In Kryex, every action on a governed system is authorized by Kryex Server against the person's live grants before it executes, not reviewed after. If the agent needs more access, it asks for it in the conversation and waits for an admin; credentials stay on the server, and every decision is written to an append-only audit log.

Kryex is built for engineering organizations that need an AI coding agent their security team can approve — especially regulated enterprises and teams that can't send code, data or credentials to someone else's cloud.

Yes. In an Enterprise deployment, Kryex Code, Kryex Server and Kryex Admin all run in your own infrastructure — in your VPC or fully air-gapped — and nothing requires Kryex Labs' servers to be reachable. Enterprise runs only the models you connect, inside your walls.

Kryex Code works with the model you choose. The free plan includes models to get started — GPT, Claude, Gemini, Qwen Coder, Mistral and more — with no API key needed. You can bring your own key for 20+ providers, point it at your company's OpenAI-compatible endpoint, or run fully local models with Ollama or LM Studio. Kryex never marks up your provider's bill.

Yes, Kryex Code is free to start, with models included and no API key needed. Enterprise — self-hosted or air-gapped deployment, Read/Write/Admin control and a customer-owned audit trail — is priced for your organization.

Kryex Code runs on Windows 10 and 11 (64-bit) and on macOS as one universal app for Apple Silicon and Intel Macs; the Mac version is an early build. There is no Linux build.

Yes. In an Enterprise deployment, IT can point every machine at your own Kryex Server through MDM — Intune, Jamf or Group Policy — and lock the agent's command policy on each device, even offline.

Admins set one command policy for every developer's agent. Command categories such as destructive file deletes, Git history rewrites, and disk and system commands can each be set to Allow, Ask or Block, and Ask and Block apply in every permission mode, Bypass included. Admins can also turn shell commands off entirely, cap the furthest permission mode anyone may pick, and lock the policy on devices, even offline.

Every database is registered once by an admin, and people get READ, WRITE or ADMIN on it. Each statement the agent runs is classified and checked against the person's live grant before it executes, so a write by someone with read-only access is blocked before it reaches the database. The agent can request more access, with a reason, and an admin decides.

No. Credentials for governed systems stay encrypted on Kryex Server, sealed per organization. Kryex Code only sends which tool to call and with which arguments, so revoking someone's access never means touching their machine.

When an admin registers an MCP server, Kryex discovers its tools and keeps every one quarantined until an admin approves it. An AI suggests a READ, WRITE or ADMIN tier for each tool, a keyword-based floor stops it understating risk, and grants are per person, per tool. If a tool's definition changes later, it goes back to quarantine until it's re-approved.

Every governed decision — access requests, grants, connections and agent tool calls — is written to a hash-chained, append-only audit log. Admins can follow it live, filter it by kind, actor, outcome, team, user or date, and export it to CSV.

People can connect GitHub, Bitbucket, Jira and Confluence, Linear, Sentry, Notion, Supabase, Asana, monday.com, HubSpot and Vercel with their own account. Admins can register PostgreSQL, MySQL, MongoDB, CockroachDB, Supabase, TiDB and MariaDB databases, Databricks and Datadog, governed web research and the agent's browser — and any other MCP server, with your own sign-in.

Kryex is built by Kryex Labs, founded in 2026 by Deepak Khimavath and Pranav Anurag.

Download Kryex Code for Windows or Mac and start on the included models — no API key needed. To run Kryex for your organization, self-hosted or air-gapped, book a demo with the team.