Quickstart

1

Enter your API key

When you open Larkup for the first time you will see the onboarding screen. Paste your API key from any supported provider, including OpenAI, Anthropic, Google, and DeepSeek, then click Save. The key is stored locally in your browser and used for chat and embedding calls.

You can change your model provider at any time from Settings → AI Models.

2

Go to the Data page and load something

Click Data in the left sidebar, then choose the source that matches your content.

• Website
• Text
• Integrations
• CSV
• PDF
• Video

Paste a page or domain URL. You can crawl it directly or search for pages by keyword.

Paste notes, transcripts, or any text you want to search, with a title.

Open Integrations, choose Notion or another source, then select Configure or Connect. After authorization, Larkup adds the selected content to your knowledge base.

Open Files and upload a CSV or Excel file. Larkup previews the columns and rows before indexing.

Open Files and drop in a PDF. You can add one file or stage several files together.

Open Media to upload a video or import a YouTube URL. Review the duration and estimated processing cost before you confirm.

Once you confirm, Larkup indexes the content in the background. You can watch progress in real time. There is no separate indexing step.

3

Open Chat and start asking

Click Chat in the sidebar. Select an LLM from the dropdown at the top and ask anything about your data. Answers are grounded in what you loaded, with source citations shown in the sidebar.

4

OR: Talk to it from social media

Connect the same Agent to Telegram, Slack, or Discord and answer questions right where your team already talks.

• Telegram
• Slack
• Discord

See Talk to Your Custom AI from Social Media to connect a channel.

Step 5: (Optional) Launch & Deploy

Customize your Agent exactly to your needs from Settings → Agent Customization:

• Prompt
• Tools
• Skills
• MCP
• Sandbox

Write the system prompt the Agent follows.

Turn built in tools on or off (e.g. semantic search, CSV analysis, form filling).

Add custom Skills to extend the Agent’s instructions and resources.

Connect a remote Model Context Protocol server.

Give your Agent access to a remote computer for code execution.

Once configured, launch your custom AI Agent server, connect to it via the SDK or REST API, and deploy it to production.

Your Agent server also exposes a complete machine readable API and interactive Swagger/Scalar reference.

See Agent Runtime API and Deployment for more details.

​

What you can do next

• Chat with a video and get answers with timestamps.
• Analyze a spreadsheet with charts and summaries in Chat.
• Sign or fill a document with a live preview.
• Talk to your Agent from Telegram, Slack, or Discord.
• Add a chat widget to your website.
• Connect an AI agent through the SDK or REST API.