Inside Atoll AI Agent: How Artificial Intelligence is Transforming Radio Network Planning

The way RF engineers interact with planning tools is about to change fundamentally. With the introduction of Atoll AI Agent, Forsk is bringing the power of Large Language Models directly into the radio network planning workflow, and the technology behind it is as elegant as the experience it delivers.

What Is the Model Context Protocol?

At the heart of Atoll AI Agent lies the Model Context Protocol (MCP), an emerging open standard that gives LLMs structured, programmatic access to external tools and APIs. Think of it as a universal translator between natural language and software capabilities.

In practice, when an Atoll AI Agent user types a request like "calculate 5G NR PDSCH CINR coverage for the current project", the following happens:

1. The LLM (ChatGPT, Claude, Mistral, Gemini, or another supported model) receives the request and reasons about the best course of action.

2. It selects from a library of 40+ Atoll API tools exposed via the MCP server.

3. The MCP server, running locally on the engineer's machine, executes the corresponding Atoll API calls.

4. Results are returned to the LLM, which synthesises them into a clear, human-readable response.

This architecture means the AI is not just generating text. It is genuinely doing things: running calculations, querying databases, creating studies, and producing reports.

Local by Design: Privacy Without Compromise

One of the most critical design decisions in Atoll AI Agent is that the MCP server runs locally. This means:

• Atoll (.atl) project files never leave the engineer's environment
• Pathloss matrices and geodata remain on-premises
• Database credentials and connection strings are never exposed to external services
  

Engineers get the intelligence of state-of-the-art AI models while their sensitive network data stays exactly where it should: on their own infrastructure.

The Skills Layer: Specialized Intelligence for RAN Planning

Beyond raw API access, Atoll AI Agent incorporates a Skills system. Skills are curated sets of instructions, scripts, and resources that the LLM loads dynamically to improve its performance on specialised tasks.

For Atoll users, this means the AI is not a generic assistant. It understands Atoll-specific workflows, MCP tool usage patterns, and RAN planning best practices. It knows, for example, that when creating a 5G NR coverage prediction, it should first check whether the predictions folder is empty, create the appropriate study type, run the calculations, and poll until completion. 

This domain-specific intelligence is what separates Atoll AI Agent from simply connecting a general-purpose chatbot to an API.

Flexibility Without Lock-In

Forsk has deliberately built Atoll AI Agent to be LLM-agnostic.

Whether your organisation standardises on ChatGPT, uses Claude for its reasoning capabilities, prefers Mistral for data sovereignty reasons, or opts for Gemini, Atoll AI Agent works with your existing AI stack. The MCP standard ensures compatibility is maintained as the LLM landscape continues to evolve.

What This Means for the Industry

Atoll AI Agent represents a meaningful architectural shift: rather than building a proprietary AI assistant from scratch, Forsk has opted for an open, extensible protocol that plugs into the broader AI ecosystem. This approach future-proofs the integration and ensures Atoll users can benefit from improvements in LLM capabilities as they emerge.