Loom / Guides
Guides
Loom works with your own keys, so the first step is getting one. These guides take you from nothing to a working chat, for every provider Loom supports.
The short version
- 1Get a key from the service you want. Each guide shows exactly where.
- 2In Loom, open Providers → Add provider, pick the service and paste the key.
- 3Choose Connect, pick a model from the chip below the message box, and start talking.
A claude.ai, ChatGPT or Gemini app subscription cannot be used, because those products have no sign-in for other apps. You need an API key, or a model running on your own computer.
Where to start
Free to try
No card needed
Start with Gemini or Groq, try free models on OpenRouter, or run a model on your own computer with Ollama.
Best quality
The big labs
Claude, GPT and Gemini are the strongest all-rounders. They bill by use, from a prepaid balance.
Most private
Stays on your network
Ollama, LM Studio or vLLM and llama.cpp keep everything on your own machine.
Start here
The providers most people begin with.
Cloud · Anthropic
Anthropic (Claude)
Claude models with an API key from the Claude Console.
Cloud · OpenAI
OpenAI (GPT)
GPT and o-series models with a key from platform.openai.com.
Cloud · Google
Google Gemini
Gemini models with a free key from Google AI Studio.
Cloud · xAI
xAI (Grok)
Grok models with a key from console.x.ai.
Cloud · Mistral
Mistral
Mistral, Codestral and more from La Plateforme.
Cloud · DeepSeek
DeepSeek
DeepSeek chat and reasoning models.
Cloud · Gateway
OpenRouter
One key for hundreds of models from many labs.
Cloud platforms
Models through your own AWS, Azure or Google Cloud account, or another platform.
Cloud · AWS
Claude on AWS
Claude Platform on AWS: an AWS-issued API key plus a workspace ID.
Cloud · AWS
Amazon Bedrock
Claude, Nova, Llama, Mistral, gpt-oss and more through your AWS account.
Cloud · Microsoft
Azure OpenAI and Foundry
GPT, DeepSeek, Grok, Llama and more deployed in your Azure resource.
Cloud · Microsoft
Claude on Azure Foundry
Claude models deployed in Microsoft Foundry.
Cloud · Google Cloud
Google Vertex AI
Gemini and partner models, and Claude, in your Google Cloud project.
Cloud · Cloudflare
Cloudflare Workers AI
Open models running on Cloudflare’s network.
Cloud · Hugging Face
Hugging Face
Thousands of open models through Inference Providers.
Cloud · NVIDIA
NVIDIA NIM
Hosted NVIDIA and open models from build.nvidia.com.
Cloud · Vercel
Vercel AI Gateway
One key, many providers, with Vercel’s routing and spend controls.
Fast and open models
Hosted open models and other model makers.
Hosted · Groq
Groq
Very fast open models.
Hosted · Together
Together AI
Open models, including Llama, Qwen and DeepSeek.
Hosted · Fireworks
Fireworks AI
Fast open models.
Hosted · Cerebras
Cerebras
Extremely fast inference for open models.
Hosted · SambaNova
SambaNova
Fast open models on SambaNova Cloud.
Hosted · Perplexity
Perplexity (Sonar)
Sonar models that search the web themselves and cite sources.
Hosted · Cohere
Cohere
Command models, with a key from dashboard.cohere.com.
Hosted · Moonshot
Moonshot (Kimi)
Kimi models. Keys are tied to the international or China site.
Hosted · Alibaba
Alibaba Qwen
Qwen models through Alibaba Cloud Model Studio (international).
Hosted · Z.ai
Z.ai (GLM)
GLM models from Z.ai.
Hosted · MiniMax
MiniMax
MiniMax models, with thinking kept out of the reply.
Hosted · Ollama
Ollama Cloud
Run larger open models on Ollama’s servers with a key.
On your own machine
Nothing leaves your network, and no key is needed.
Anything else
Any service that speaks the OpenAI API.
Beyond models
Search, connectors and fixing problems.
No guide matches that. Try a shorter word, or use Custom for any OpenAI-compatible service.