Skip to content

Documentation

From install to shipping, in five minutes.

Clear, practical guides — open a folder, write code into files from the AI panel, run the agent, flip on Sovereign mode, and activate your license. Jump to a topic below.

Install

Native Windows app, offline-ready.

Open a folder

Any local directory becomes your workspace.

Press ⌘K

Describe a change, accept the inline diff.

5 minutes

Getting started

Install Obsydian Lab, launch it, and start your free trial — no card, no setup wizard to fight.

  1. 1

    Download & install

    Grab the Windows installer from the download page and run it. Obsydian Lab is a native desktop app built on a Tauri shell, so it launches instantly and runs offline.

  2. 2

    Start your trial

    The first launch begins a 14-day free trial with no card required. Every feature — the full editor, the agent, included AI — is unlocked from minute one.

  3. 3

    Make your first edit

    Open any file, select a block, press ⌘K, and describe the change in plain English. Accept the inline diff. That single loop is the fastest way to feel what the editor does.

No API keys needed — frontier AI is built in. Just open the editor and ask.

Your project

Opening a folder

Point Obsydian Lab at any local directory and let it index your project so AI features understand your code.

  1. 1

    Open a folder

    Use File → Open Folder (or the Explorer's open button) to choose any directory on your machine. It becomes your workspace — files stay exactly where they are on disk.

  2. 2

    Let it index

    Obsydian Lab builds a semantic @codebase index in the background. This is what lets the AI reason about your project as a whole rather than one file at a time.

  3. 3

    Navigate fast

    Press ⌘P to jump to any file by name, and use the Explorer right-click menu to create, rename, or move files. The visual git graph shows branches and history without leaving the editor.

Write-into-file

Using the AI panel

Ask in plain English and let the AI write directly into your files — with a reviewable diff before anything lands.

  1. 1

    Open the panel

    The AI panel sits beside your editor. Choose a mode — Ask to discuss, Edit to change the current file, or Build to scaffold something new — then type your request.

  2. 2

    Write into a file

    In Edit mode, describe the change and the AI rewrites the relevant code directly into your file. You review the inline diff and click Accept or Reject — nothing is written until you say so.

  3. 3

    Pick the right model

    12+ providers are built in. Route a task to Claude (the default coding brain), GPT, Gemini, DeepSeek and more from the model picker. Switch mid-task and your context carries over.

Use @file to pull a specific file into context, or @codebase to ground a request in your whole project.

Hands-off

Agent mode & Composer

Hand off a goal and let Composer plan, edit across files, run tests, and hand back a single reviewable diff.

  1. 1

    Describe an outcome

    Switch to Build/Agent mode and state what you want — "add pagination to the users table," not step-by-step instructions. Composer drafts a plan before touching a file.

  2. 2

    Watch it work

    Composer edits across the relevant files, runs your test suite, and iterates. It's grounded by the @codebase index, so its edits respect your existing patterns and conventions.

  3. 3

    Approve the diff

    Every write to disk and every shell command passes through an approval gate — nothing risky happens silently. Review the final diff, then accept, reject, or refine.

Keep the agent on a tight leash while you build trust, then loosen approvals for routine work.

Offline-first

Sovereign mode

One switch keeps every byte on your machine — local models only, no cloud, no telemetry, even with the network unplugged.

  1. 1

    Flip the switch

    Open Settings → Privacy and enable Sovereign Mode. The editor immediately collapses every third-party network surface to local-only.

  2. 2

    Point it at Ollama

    Install Ollama and pull a model. With Sovereign Mode on, chat, edits and autocomplete all route to your local models — the editor works fully offline.

  3. 3

    Know what stays

    Your files are never uploaded for indexing or training. In Sovereign Mode, nothing phones home — no update checks, no search, no telemetry. Privacy is the foundation, not a setting.

Billing

Activating a license

Move from trial to a paid plan and activate your license — online or with an offline key.

  1. 1

    Choose a plan

    When the trial ends, pick Pro ($20/mo), Max ($100/mo), or per-seat Team ($100/seat/mo) from the pricing page. Every plan includes frontier AI with no separate token bill.

  2. 2

    Activate your key

    Paste your license key in Settings → Account, or sign in to activate automatically. Offline keys keep the editor working without a constant connection.

  3. 3

    Manage or cancel

    Update payment details, change plans, or cancel anytime from the self-serve billing portal. Your in-app quota meter shows exactly how much included AI you've used this month.

Sovereign by design

Your code never leaves your machine.

Everything in these guides runs on a local-first foundation. Files stay on your disk, and Sovereign Mode processes every byte on-device — no cloud, no telemetry, no exceptions.

Go deeper

A full IDE, not a chat box.

AI & providers

Route each task to the best of 12+ models — Claude, GPT, Gemini and more.

Terminal & git

A streaming integrated terminal and a visual git graph, built in.

Debugger & LSP

Real DAP breakpoints, stepping, diagnostics and go-to-definition.

Best learned by doing.

Install the editor and start the free trial — most things click in the first five minutes.

Download for Windows Explore features