Connecting Blocks agents to Replit

Creating AI agents is one thing, but often you need frontend if you want to make it easy for others to use it. That’s where no-code platforms like Replit come in. Replit is a straightforward place to build that page. A skill file tells Replit how to call the Blocks agent, so the connection does not have to be rewritten for every app.

To see this in action, check out this LinkedIn Company Report. The page calls the LinkedIn company report agent on Blocks. That agent reads the company’s public LinkedIn activity and returns a short report you can read on the page and download as Markdown.

Let’s go through the quick steps you can take to build this yourself.

Prerequisite: A Blocks.ai account and API key

Before getting started, make sure you have set up an account at Blocks.ai. Once you are logged in, you will need an API key — this is what identifies you when you call a Blocks agent. You can create one on the API keys page.

Step 1: Add the skill to your Replit workspace

You need to let Replit know how it can call any Blocks agent. To do this, you’ll add a skill. Go to Workspace Settings → Customization → Skills. Then add a new skill, select “Add from GitHub” and provide this URL:

Blocks caller skill

You should then see a ‘blocks-caller’ skill added as a customization, that you can use in any chat by typing ‘/blocks-caller’.

Step 2: Prompt a new app

Open a new Replit window. In Agent chat, select the skill from the dropdown, then send a prompt like this explaining briefly what you want to build:

I want to setup a frontend for this agent: https://app.blocks.ai/agents/linkedin_company_report_agent It should take in the name of a company, and then format the outputted md file, with an option to download.

After the prompt, type ‘/blocks-caller’ and you should see the skill added as bubble text below the prompt.

Step 3: Review the page

Look at the page Replit builds. Ask Replit to change the layout, wording, or download until it matches what you want.

Step 4: Try this yourself with other agents

You can check out this list of public Blocks.ai agents and try building a frontend for any of these. Just change the URL and any input/output details in the prompt above.