Want to use a model of your own choosing with Now Assist – instead of one of ServiceNow's managed providers? Maybe it's a model you've fine-tuned, an open-weights model you're hosting yourself, a provider ServiceNow doesn't have a relationship with, or a data-residency requirement that means nothing can leave your network. Whatever the reason, this video shows you how to connect it. We set up Claude end to end in ServiceNow's Generic LLM Connector, then connect Gemini, Hugging Face, DeepSeek and the brand-new Moonshot Kimi K3 using the same pattern. We also test the result against a real custom Now Assist skill, not just the setup wizard's own connection test. *You'll learn:* - Why an organisation might connect its own LLM instead of using a ServiceNow-managed provider - How AI Control Tower governs whether custom LLMs are permitted on your instance in the first place - How to connect a custom LLM end, demonstrated step by step with Claude - How the same pattern connects Gemini, Hugging Face, DeepSeek, and Moonshot Kimi K3 - Credential fields the setup wizard doesn't ask for, and where to set them - How to test a custom LLM against a real custom Now Assist skill - Why fetching the credential inside the request transformer with ConnectionInfoProvider passes the connection test and then breaks once role restrictions are applied from the skill *Chapters:* - 0:00 Intro - 1:35 AI Control Tower - 2:40 Overview of connecting to an LLM - 3:46 Connecting a model (Claude) - 7:55 Fixing the credential - 11:25 Gemini and Hugging Face - 12:52 Kimi K3 and DeepSeek - 14:05 Testing from a real skill - 15:37 Authenticate the right way - 17:10 Outro *Links:* - Sample transformer scripts for all five providers (GitHub) – https://github.com/jnichols-servicenow/LLM-Demo-Transformer-Scripts - ServiceNow: Generative AI Controller – https://www.servicenow.com/docs/r/intelligent-experiences/generative-ai-controller/generative-ai-controller.html - ServiceNow: Create a model – https://www.servicenow.com/docs/r/intelligent-experiences/create-model.html - ServiceNow: Configure a Generic LLM Connector – https://www.servicenow.com/docs/r/intelligent-experiences/configure-a-generic-llm-connector.html - ServiceNow: A2A API Key credential behavior – https://www.servicenow.com/docs/r/intelligent-experiences/a2a-api-key-credential-behavior.html - ServiceNow Community: Using external LLMs with Now Assist (including BYOK/BYOLLM) – https://www.servicenow.com/community/now-assist-articles/using-external-llms-with-now-assist-including-byok-byollm/ta-p/3218103 - Anthropic: Messages API reference – https://platform.claude.com/docs/en/api/messages - Google Gemini: Interactions API reference – https://ai.google.dev/gemini-api/docs/interactions-overview - Hugging Face: Inference Providers – Chat Completion – https://huggingface.co/docs/inference-providers/en/tasks/chat-completion - DeepSeek: Create Chat Completion – https://api-docs.deepseek.com/api/create-chat-completion/ - Moonshot AI (Kimi): Create Chat Completion – https://platform.moonshot.ai/docs/api/chat - RFC 9110 §11 — HTTP Authentication (the Authorization header and the auth scheme the Bearer prefix belongs to) — https://www.rfc-editor.org/rfc/rfc9110.html#section-11 - RFC 6750 — The OAuth 2.0 Authorization Framework: Bearer Token Usage — https://www.rfc-editor.org/rfc/rfc6750.html *Endpoints used in this video:* - https://api.anthropic.com/v1/messages - https://generativelanguage.googleapis.com/v1beta/interactions - https://router.huggingface.co/v1/chat/completions - https://api.deepseek.com/chat/completions - https://api.moonshot.ai/v1/chat/completions