Bring Your Own Key

Connect your own provider for unlimited AI usage on any plan, including the free tier. You pay your provider directly and Boltise never meters the usage.

Caution

One trade-off worth knowing before you choose. Hosted Boltise AI retrieves from a FiveM/RedM knowledge base on our servers before the request reaches the model. With your own key the request goes straight from Boltise to your provider, so that retrieval step does not happen and the knowledge base is not in the prompt.

You keep everything else — the full tool surface, project context, approvals, and the built-in natives index. But on framework questions specifically, hosted answers are better grounded.

Supported providers

Provider	Models	Get a key
Anthropic	Claude Opus, Sonnet, Haiku	console.anthropic.com
OpenAI	GPT-5.5, GPT-5.4 and Mini/Nano, Codex	platform.openai.com
Google	Gemini 3 Pro, 2.5 Pro, 2.5 Flash	aistudio.google.com
OpenRouter	200+ models	openrouter.ai
DeepSeek	DeepSeek Chat	platform.deepseek.com
Groq	Llama 3.3 70B, Mixtral 8x7B	console.groq.com
Mistral	Mistral Large, Small	console.mistral.ai
xAI	Grok 3, Grok 3 Mini	console.x.ai
Kimi (Moonshot)	Kimi K2, Moonshot v1 128K	platform.moonshot.cn
Together AI	any model you name	api.together.xyz
Fireworks AI	any model you name	fireworks.ai
Ollama	local models, no key needed	ollama.com

Together AI and Fireworks AI are model hosts — they serve other vendors’ models under ids that change as their catalogues do, so Boltise does not ship a fixed list for them. Type the model id your account has access to, the same way you do for Ollama.

Kimi runs two separate platforms — .cn (China) and .ai (international) — with separate accounts and non-interchangeable keys. Boltise defaults to the .cn endpoint.

Set up a key

1. Open Settings ▸ AI Providers (or click Configure providers at the bottom of the model picker in the chat panel).
2. Find your provider and paste your API key.
3. Back in the chat panel, open the model picker and choose a model under that provider.

Your key is stored locally and is never sent to Boltise’s servers.

Ollama (local, private)

Run models on your own machine, free and offline.

1. Install Ollama.
2. Pull a model: ollama pull llama3.2.
3. Select Ollama in Boltise.
4. Confirm the URL (http://localhost:11434 by default) and pick your model.

Note

Local models are also the slowest and least capable option here. For FiveM/RedM answers, a hosted cloud model — yours or Boltise’s — will do noticeably better.

OpenRouter

1. Get a key from openrouter.ai.
2. Select OpenRouter in Boltise.
3. Enter the key and pick from 200+ models.