SurfacesAvailable

Fix the finding in code and run the check.

Dex, working through a task

Deixic Code is a native coding agent for the terminal, scripts, and CI. It changes the code behind a failing security finding, runs your tests before you commit, and asks you before it runs a risky command. Settings in your other systems are changed through Deixic's connections, with a receipt.

curl -fsSL https://raw.githubusercontent.com/dx-corp/code/main/scripts/install.sh | bash

Request early access

A code-side finding, fixed and checked.

Example terminal session with demo data.

  1. Finding

    A scan flags a database password committed in a deploy config file.

  2. Change

    Dex moves the value to an environment reference and removes it from the file.

  3. Check

    Dex runs your test suite before the commit. 214 tests pass.

  4. Command approval

    Dex asks before running a history rewrite to purge the old value. You approve or deny in the terminal.

  • Terminal, scripts, CI

    Run it interactively, from scripts with exec, or headless in CI so the check runs on every change.

  • Guardrails on your machine

    An action firewall and approvals gate risky commands, with an optional macOS or Linux sandbox, signed releases, and automatic updates.

  • Commands cannot sneak past

    The firewall reads commands the way your shell does, so wrappers, quoting, and escapes do not hide a dangerous one.

  • A second model checks the work

    Pair a primary model with a reviewer model from a different provider.

  • A team of agents

    Split big changes across explore, plan, code, and review agents, each in its own worktree.

  • Undo a wrong turn

    Rewind the conversation to an earlier turn and, in a Git repository, the file changes with it.

  • Pick up where you left off

    Resume, branch, or move a session to another machine with credentials stripped out.

  • In your editor

    Runs inside IntelliJ, PyCharm, the JetBrains family, Zed, and any editor that speaks the Agent Client Protocol.

Deixic Code reading and updating SECURITY.md after approval, then running two passing local regression tests.
The native terminal UI with demo data. A local security-policy fix is tested; production publication and the control re-test remain pending.

All surfaces.