Tenixe AI

Chat, a coding agent and image generation on your Tenixe plan, inside Visual Studio Code.

What it does

• Sign in with your Tenixe account: VS Code shows a short code, you approve it on tenixe.com, and you are in. No password is typed into VS Code.
• Chat in Ask and Plan mode with the models on your plan. Attach the open file, your selection (Ctrl+Alt+A) or other files from the folder (type @); copy or insert code from an answer at your cursor; stop an answer at any time.
  • Each answer shows the model, how long it thought, its tokens and how long it took, with Copy and Retry; code is coloured; the files it names are listed to open or attach; three follow-up questions under it put the next question in the composer in one click.
  • The context window ring in the composer shows how full the chat is, with the breakdown and the tokens billed in this chat — and Compact now (summarise the older messages) or New chat with summary when it fills up.
  • Chats keep working in the background: start a new chat or open another while one answers or runs. The chat list marks them — blue while working, amber when one waits for you.
  • Open in editor moves the chat to an editor tab; the sidebar then lists your chats. Move to sidebar brings it back.
• Agent — switch the composer to Agent (or run Tenixe AI: Run Agent…) and give it a task. It reads, searches and edits files and runs commands on your computer, in the open folder, and shows every step as it goes. The ⓘ beside the mode says what it may do here and what it does without asking.
  • It asks before it runs a command or makes a risky change, showing the exact command. "Always allow in this chat" covers that exact command only.
  • If a file it is about to change has unsaved edits in your editor, they are saved first, so nothing you typed is overwritten.
  • What it changed, line by line: under each edit, the removed (red) and added (green) lines with their numbers; a changed file unfolds to its whole diff, and a click opens the file at that line.
  • Review in the editor, as in TX Code: the agent's lines are highlighted in the file, with Keep / Undo over each change, and Keep / Undo for the whole file in the editor's title bar.
  • Every changed file can be opened in a diff against what it was before the run, and reverted — one file or all of them.
  • Stop ends the run at once. If the window closes or the connection drops, the run waits on the server and shows up under Unfinished agent runs, ready to resume.
  • It understands your code the way VS Code does: go to definition, references, hover, outline and problems come from the language extensions you have installed, so it follows imports instead of guessing from text.
  • A command it starts in the background (a dev server, a watcher) gets its own read-only terminal; Ctrl+C there stops it.
  • A web page it builds opens in a preview with no network access.
  • Project hooks in .txcode/hooks.json run only after you have seen and allowed every command they contain — and again after any edit to the file.
  • The agent runs one task at a time per window; the composer of another chat says which chat has it.
• In the editor
  • Suggestions as you type — grey text at the cursor; Tab accepts. Turn them off or on from the status bar (Tab) or the setting.
  • Edit a selection — Ctrl+Alt+K (Cmd+Alt+K on a Mac): say what to change; with nothing selected it edits the current line. You see a diff and decide with Accept / Reject above the change.
  • Apply on a code block in an answer puts it into the right file — the file the answer named, or the one you pick. Code that is only part of a file is merged into it. You always see the diff first, and nothing is written until you accept; Accept is an ordinary edit, so Undo works.
  • Fix / Explain with Tenixe AI in the lightbulb of an error or warning.
  • When a command fails in a terminal, Ask Tenixe AI offers to look at it (needs VS Code's shell integration).
  • These put a ready-made message in the composer for you to read and send; nothing is sent on its own.
• Pictures — ask for one in the chat ("design an app icon"), in any mode: the agent makes it with your plan, shows it in the chat and saves it into your project — Open, Insert path, Copy path. The gear in the composer opens Image settings: the model and quality (shared with TX Code), shape, pictures per request, the price, and where new pictures go — let the AI pick, a folder you choose, or only your gallery on tenixe.com. Under the pictures: More like #N, Redo in High quality and Set as app icon with Agent.
• Problems — attach the active file's errors and warnings to a message (⚠ in the composer, or Add Problems to Chat).
• Several folders open? Pick the one Tenixe AI works on in the composer.
• History — reopen and continue the conversations of the open folder, searched and grouped by day.
• Account — your plan, how much of each quota is used (a percentage) and your credits at a glance.

Keyboard shortcuts

They avoid VS Code's own Ctrl+K chords; change them in Keyboard Shortcuts.

Getting started

1. Open the Tenixe AI view in the activity bar (Ctrl+Alt+L).
2. Choose Sign in, then Open browser, and approve the code on tenixe.com.
3. Open a folder and ask a question.

The walkthrough Get started with Tenixe AI (Help → Welcome → Walkthroughs) goes through the same steps, then the agent, the editor tools and images.

It uses the same Tenixe account, models, quota and credits as TX Code and tenixe.com. Nothing is billed outside your plan.

Settings for the agent

Settings for the editor

Reverting puts back the content a file had before the run; it is not VS Code's undo, and edits you made to that file afterwards are lost (you are asked first). Use git for anything you want to keep for sure.

Requirements

• Visual Studio Code 1.96 or newer.
• A Tenixe account (tenixe.com) on a plan that includes Tenixe AI for VS Code.

Privacy

What you send in the chat, the files and command output the agent reads while working on your task, and — while suggestions are on — the code around your cursor, are sent to Tenixe's servers and to the AI model provider your plan uses, to produce the answer. Terminal output is read only to decide whether to offer help, and leaves your machine only in a message you send.

The extension collects no telemetry of its own, with one exception: while suggestions are on and VS Code's telemetry is enabled (telemetry.telemetryLevel), it tells Tenixe whether each suggestion was accepted — an id, yes or no, how long it was shown and the model, never your code — to measure their quality. See the Privacy Policy and the Terms of Service.

The assistant stays off in folders VS Code has not trusted (Restricted Mode), because it reads and edits files and can run commands.

Support and feedback

tenixe.com/dashboard/support — also Help & feedback in the Account tab. Please include the version (bottom of the Account tab) and, for a problem, the output of Tenixe AI: Show Logs.