ambleloftGet started with Ambleloft
Start the Ambleloft desktop AI agent, connect a model and try a file task. Add extensions when you need more capabilities.
Outcome: Start the agent, connect a model and review your first task output.
Before you begin: Start with the environment setup below; no extension development is required.
On this page
01Prepare your environment
Use Node.js 22.13 or newer and npm. The steps below run the public repository source. New collaboration features require the corresponding source revision; unpublished working-tree changes may not yet be merged. Check release notes for installer coverage and validate your target platform.
02Run from source
Clone the repository for a first setup. In an existing checkout, use the install, engine preparation and startup steps. Engine preparation downloads the pinned runtime; a separate Codex CLI installation is not required.
git clone https://github.com/fjb040911/Ambleloft.git
cd Ambleloft
npm ci
npm run engine:prepare
npm run dev03Connect a model
Open Settings → Model providers → Add service. Enter the Base URL, model ID and API key, choose the protocol, then test. Do not append /responses or /chat/completions to the Base URL. Connection tests may incur provider charges.
04Try your first agent task
Create a task, attach a sample project description and ask: Organize this document into completed work and items to check. Use a model that supports tool calling. Follow the work, answer questions or confirm actions as needed, then review the output. You do not need to develop an extension for this step.
05Optional: run the expense extension demo
Build and install the expense sample, start its demo service and configure the URL as its README describes. Ask to claim travel expenses in chat, then enter details, confirm and review the result. Your model must support compatible tool calling.
Open the complete sample and instructions06Check results and recover
Confirm the desktop window opens and the configured model answers the task. If engine preparation or startup fails, keep the terminal log and check downloads, credentials and occupied ports. Successful dependency installation does not verify the desktop or model.
Check the result and troubleshoot
Start the agent, connect a model and review your first task output.
Check actual results using this page’s verification steps. For business integrations, inspect UI submission, tool receipts and server records separately.
Open the error reference →