🤖 Arc Assistant

The Arc Assistant is an AI assistant built into Arc that helps with TM1 development. It can answer questions about your model, draft and validate MDX and TI, build cube views, and, when you allow it, read and write cell data through a set of assistant tools. It works with OpenAI-compatible providers as well as Ollama, Hugging Face, and WatsonX, and can be extended with external Model Context Protocol (MCP) servers.

This section covers how to enable the assistant, control what it is allowed to do, customise its behaviour, and connect it to external tools. Every option described here is set in settings.yml in the Arc installation directory. For a complete parameter list, see the Arc settings.yml File Reference.

Getting started

• Enable the Arc Assistant — turn the assistant on and configure the AI service provider, model, and API key.

• Disable the Arc Assistant — turn the assistant off entirely.

Governance and security

• Controlling Arc Assistant Automation and Tool Limits — disable tool calling for conversational-only use, and cap how many tool calls the assistant may make in a single turn.

• Restricting Arc Assistant Access to TM1 Objects and Data — limit which objects the assistant can discover, cap cube view reads, and redact sensitive cell values.

Customisation

• Customising Arc Assistant Prompts and MDX Templates — add per-model prompt instructions and supply your own MDX templates.

• Generating Unit Tests with the Arc Assistant — configure AI-enhanced unit test generation in the Test Hub.

External tools (MCP)

• Enabling MCP Servers for the Arc Assistant — declare external MCP servers in settings.yml, including a Pulse example.

• Connecting an MCP Servers to the Arc Assistant — the follow-on step of connecting to a configured MCP server from the Arc user preferences.