Tsedal AI Get early access
Private beta Now accepting developers

Every frontier model.
One extension.

Claude, GPT, Gemini and DeepSeek inside VS Code. Ask a question, or hand a whole task to the agent.

Download for VS Code Get early access

Free during the private beta · No credit card · Works in VS Code

orders.py · shop-api
orders.pymodels.pytest_orders.py
Tab to accept
main Tsedal AI · Ready

Powered by frontier models from the world's leading AI labs

Features

Ask a question. Or hand off the whole task.

Two ways to work with Tsedal: an agent that does the work, and an assistant that answers. Both inside the editor you already use.

Agent mode

An agent that ships the change

Describe what you want built or fixed. The agent reads your code, plans the change, edits across files, runs commands and tests, and keeps going until the job is done. You review every change before it lands.

  • Works across your whole project, not one file
  • Runs terminal commands and tests, with your approval
  • Every edit shown as a diff you can accept or reject
Ask mode

Answers about your code, without touching it

Ask about the file in front of you, the lines you've selected or an error you're stuck on. Ask mode only reads, so it never changes your files. Answers stream in as they're written.

  • Explain unfamiliar code
  • Track down bugs from a stack trace
  • Plan a change before you make it

Scheduled tasks

Give the agent a task to run at a set time or on a repeat, like a morning dependency check or weekly release notes, and come back to the results.

Attach files

Drop in screenshots, specs, logs or any file, and Tsedal works with them as context, in Ask or Agent mode.

Edit with a sentence

Select code, describe the change, and review the diff before anything is applied.

Autocomplete that reads ahead

Suggestions as you type, written for the code before and after your cursor. Press Tab to accept.

Intelligent routing

The right model, every time.

Quick questions get instant answers. Hard problems get deep reasoning, automatically. Try it: this runs the same rules Tsedal uses, right in your browser.

Level
You Tsedal Fast
Answered by Fast

Difficulty

    Intelligence levels

    From instant answers to frontier reasoning.

    Pick a level when you want control, or let Tsedal choose for each request, in Ask and Agent mode alike. Small chores like commit messages go to a lightweight model.

    Fast

    Your everyday pair programmer. Answers in a second or two.

    • Chat and inline edits
    • Bug hunting
    • Moves up to Smart when a question looks hard
    SpeedDepth
    Recommended for hard problems

    Smart

    Thinks through the problem before it answers.

    • Tricky bugs and stack traces
    • Refactoring and design
    • Chosen for you when it matters
    SpeedDepth

    Max

    The most capable frontier model, for the hardest problem of your day.

    • Architecture decisions
    • Complex algorithms
    • Large, subtle changes
    SpeedDepth

    Autocomplete

    Always on. Suggestions as you type, tuned for speed.

    • Fills in the middle of your code
    • Lines and whole blocks
    • Tab to accept
    SpeedDepth
    Added to each request by routing
    <50ms
    Lines of your code we store
    0
    To switch models if one stalls
    8s
    Leading AI labs behind it
    6

    If a model is slow or unavailable, Tsedal moves your request to the next best one automatically, so you keep working.

    Privacy

    Private by design.

    Your code is yours. Tsedal keeps a record that a request happened, never what was in it.

    Read our terms and privacy

    Never stored

    Prompts, code and answers are never saved by Tsedal. We only count usage.

    Encrypted in transit

    Everything between VS Code and Tsedal travels over an encrypted connection.

    You stay in control

    Choose a mode yourself whenever you like. Tsedal never overrides your choice.

    Pricing

    Free while we're in private beta.

    Join now and use every mode, including frontier models, at no cost during the beta.

    Private beta
    $0during the beta

    No credit card. No payment details. Just your email.

    Get early access
    • Agent and Ask modes in VS Code
    • Scheduled tasks and file attachments
    • Fast, Smart and Max levels, including frontier models
    • Automatic routing to the right model
    • Generous daily allowances, always visible in your status bar
    • Your code is never stored
    • Direct line to the team: your feedback shapes the product

    FAQ

    Questions, answered.

    Can't find what you need? Ask us when you join the beta.

    Is Tsedal AI free?

    Yes, during the private beta. Every account gets generous daily allowances for each mode, with no credit card or payment details. Your status bar always shows what you have left.

    What's the difference between Agent and Ask mode?

    Ask mode answers questions about your code and never changes your files. Agent mode carries out a whole task: it reads your project, edits files, runs commands and tests, and keeps going until it's done. Switch between them at any time in the Tsedal panel.

    Is it safe to let the agent change my code?

    Every change the agent makes is shown as a diff you can accept or reject, and you decide which actions, such as running terminal commands, need your approval first. Use version control as usual so you can always go back.

    Which AI models does Tsedal use?

    Frontier models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), DeepSeek, Moonshot AI (Kimi) and MiniMax. Tsedal matches each request to the model best suited to it, and the lineup keeps improving as new models are released.

    Do you store my code?

    No. Tsedal never saves your prompts, code or answers; it only records that a request happened and how large it was. Each request is processed by the model that answers it, as described in our terms.

    What happens when I use up my allowance?

    You'll see a clear message telling you when it resets, and you can keep working in another mode in the meantime.

    What's the difference between Smart and Max?

    They are intelligence levels, available in both Ask and Agent mode. Smart thinks a problem through before answering and suits tricky bugs and refactoring. Max uses the most capable frontier model available, for the hardest problems. In Auto, requests that look hard are moved up to Smart for you.

    Which editors are supported?

    Visual Studio Code, through the Tsedal AI extension. An internet connection is needed while you use it.

    Private beta

    Start coding with Tsedal AI.

    We're welcoming developers in small groups. Leave your details and we'll send your access and the VS Code extension as soon as your place is ready.

    By requesting access you agree to our terms of use and privacy.