How your AI finds the food

Traditional SEO is about ranking high on Google. We're building something different: structured data that AI assistants understand directly.

The problem with Google search

When you search for "local food in Oslo" on Google, you get ads, big grocery chains, and maybe a blog post. The small producers drown.

AI assistants work differently. They don't read websites — they fetch structured data from protocols designed for machine-to-machine communication.

🔍 Traditional search

  • Based on website ranking
  • Favours large players with SEO budgets
  • Ads dominate the results
  • Text designed for humans

🤖 AI-driven search

  • Based on structured, verified data
  • Equal terms for all producers
  • No ads in the results
  • Data designed for machines

The protocols we use

Rett fra Bonden uses open standards that let any AI assistant find and understand information about Norwegian matprodusenter:

🌐

A2A

Google's Agent-to-Agent protocol. Agents communicate directly with each other.

🔌

MCP

Anthropic's Model Context Protocol. Claude and other AIs fetch data as tools.

📖

Schema.org

Structured markup that Google Rich Results understands.

🛠

OpenAPI

Standard API specification. Any developer can integrate.

How it works in practice

Here's an example of what happens when you ask an AI assistant "where can I find fresh vegetables in Oslo?":

// 1. The AI assistant sends a request via the A2A protocol
POST rettfrabonden.com/api/a2a

// 2. Our agent finds relevant producers
{ "query": "vegetables oslo", "results": [...] }

// 3. Structured data returns with opening hours, contact info, certifications
{ "name": "Grønn Bonde", "hours": "Mon–Sat 08–16" }

Everything happens automatically. The producer doesn't have to do anything — we collect data from public sources, verify it, and make it available to all AI platforms.

Set up MCP — search from your AI

MCP (Model Context Protocol) lets your AI assistant search our database of 1405+ matprodusenter directly. Here's how to set it up:

💚 ChatGPT (easiest)

1
Go to chatgpt.com and open a new conversation
2
Click the tools icon (🔧) in the message field and choose "Add an MCP Server"
3
Paste this URL: https://rettfrabonden.com/mcp
4
Done! Try for example "Find organic honey in Bergen"

🧰 Claude Desktop (Pro/Max/Team/Enterprise)

Method 1 — Remote MCP (recommended, no install):

1
Open Claude Desktop → Settings → Integrations
2
Click "Add custom connector"
3
Paste: https://rettfrabonden.com/mcp
4
Done! Try for example "Find organic meat in Trondheim"

Method 2 — Local npm package (for developers, Claude Code, or Claude Desktop without Pro):

1
Install Node.js → Open Claude Desktop → Settings → Developer → Edit Config
2
Add:
{
  "mcpServers": {
    "lokal": {
      "command": "npx",
      "args": ["lokal-mcp"]
    }
  }
}
3
Save (Ctrl+S) and restart Claude Desktop.

⚙️ Other AI platforms

Any platform that supports MCP Streamable HTTP can connect: https://rettfrabonden.com/mcp

For REST-based integrations, see our OpenAPI specification.

Open source

The whole project is open source. We believe infrastructure for food visibility should be a shared good, not a commercial product.

See the code on GitHub →