Models & Providers
MeetLoyd doesn't lock you into one AI vendor. You bring your own keys and choose the best model for each job — across ten providers, plus self-hosted open-source models for full sovereignty. Your agents, your models, your bill.
"Which model should this agent use?" or "Add our OpenAI key and let the team use GPT." Loyd can connect a provider and help you pick the right model for the task. Meet Loyd →
Bring Your Own Keys (BYOK)
BYOK is the default at every tier. You connect your own provider API keys, and your usage is billed by the provider, to you — MeetLoyd doesn't sit in the middle adding a markup on your tokens. Benefits:
- No vendor lock-in — switch models per agent, per task, any time.
- Your commercial terms — you keep your own provider pricing and rate limits.
- Your data path — calls go from MeetLoyd to the provider you authorized.
Keys are stored encrypted and are revocable at any time.
Ten providers, one platform
Pick whichever model fits the task — quality, speed, cost, or data residency:
| Provider | Models | Good for |
|---|---|---|
| Anthropic | Claude family | High-quality reasoning, long context (default) |
| OpenAI | GPT family | General-purpose, broad ecosystem |
| Gemini family | Multimodal, long context | |
| Mistral | Mistral / Codestral | Efficient European models, coding |
| Groq | Llama family | Very fast inference |
| DeepSeek | DeepSeek | Strong reasoning at low cost |
| Alibaba Qwen | Qwen | Capable multilingual models |
| Zhipu | GLM | Capable general models |
| Moonshot | Kimi | Long-context models |
| vLLM (self-hosted) | Open-source (DeepSeek, Qwen, GLM, Llama Guard, …) | Full sovereignty — runs on your own infrastructure |
Microsoft Azure OpenAI is also supported, including multiple deployments under a single connection — the shape regulated Microsoft 365 customers need.
Always-current model lists
You don't have to wait for us to add a model. When you connect a key, MeetLoyd asks the provider directly which models that key can access and offers exactly those — so a newly released model shows up the moment your key can reach it. Model aliases (like "latest") keep your agents on the current version without manual updates.
Sovereignty: self-hosted open models
For the strictest data requirements, MeetLoyd runs open-source models on your own infrastructure via vLLM. Nothing leaves your perimeter, and you still get the same agents, governance, and audit trail. This pairs with two related controls:
- Data sovereignty — keep your business data under your own keys (CMEK) or in your own storage (BYOS), so both your data and your inference stay yours.
- Sovereign moderation — content safety runs on a self-hosted model too, so even moderation never sends your content to a third party.
Your CISO picks the menu
Model choice is governable. An administrator can set a tenant-wide approved-model list and run it in audit, warn, or enforce mode — so agents can only use models your organization has sanctioned, enforced on every run and recorded in the audit trail. The platform only offers models your keys can actually reach, so the menu reflects reality.
What you get
- The right model for each job — not one vendor's answer to everything.
- No token markup — BYOK means you pay your provider directly.
- No staleness — new models appear as soon as your key supports them.
- Sovereign option — open models on your own infrastructure when you need it.