Lazy Coding Agent
Menu
Download

Documentation → Getting started

Getting started

Lazy Coding Agent is a desktop AI coding workspace. You open a project folder as a workspace, chat with an LLM in multiple sessions, and (in Agent mode) let the assistant read and change files under that folder using tools.

On launch, a branded splash image appears while the workspace starts (about one second after loading finishes). Window chrome follows your saved theme (dark by default, or light if you chose View → Toggle Light / Dark Theme last time).

Open or create a workspace

  1. Use File → Open Project Folder to point at an existing code tree, or create a workspace from the home flow when offered.
  2. The status bar and chat hero show the active workspace name.
  3. Panel layout, theme, and dock state are saved per workspace in application settings.

Settings you need first

Open Settings from the menu (or when prompted):

  • LLM provider — API key, base URL if needed, and default model.
  • Fetch Models — loads model IDs from your provider (requires working HTTPS).
  • Agent Behavior — optional Debugging mode to show full tool JSON in chat.

Save settings before starting a long agent run.

Building from source (developers)

Install a release from Download or build from the application repository. The README documents Lazarus package install, lazbuild, and post-build steps that copy help/ and assets next to the executable so Help → User Guide works outside the IDE.

HTTPS and OpenSSL (Linux / Raspberry Pi)

Chat, Fetch Models, and the agent need HTTPS. On Linux the app uses the system OpenSSL 3 libraries.

  • arm64 .deb (Raspberry Pi OS, Debian, Ubuntu): Installing the release package should pull in libssl3 automatically. If HTTPS still fails after an upgrade, run sudo apt install libssl3 (on some distros the package is named libssl3t64).
  • Built from source on the Pi: Install the runtime before using AI features: sudo apt install libssl3. You do not need the -dev package to run the app.

Settings and chat show a clear error instead of crashing when OpenSSL is missing.

Your first chat

  1. In the left rail, open Agents and choose + New Chat Session (or use + on the chat session tab strip).
  2. Pick a mode in the composer footer (Ask, Plan, or Agent).
  3. Choose a model from the combo beside it.
  4. Type a prompt and send.

Use Ask for questions without file writes. Use Agent when you want the assistant to edit the project.

Where to learn more