An AI platform with hosted or BYOK billing: hosted tiers bundle token budgets across 10 provider catalogs; BYOK tiers use your organization keys so you can use your own OpenAI or Anthropic key and pay vendors directly. The AI control plane for LLM keys stays the same—agents, retrieval, and channels do not change when you switch economics.
After sign-in, add or rotate credentials under Settings → LLM API Keys.
Hosted plans
Hosted tiers (free through enterprise) map to monthly token allowances and model tiers. Heavier frontier models draw down the same budget faster—so you can mix models while staying inside plan economics.
Each hosted tier ships with a defined monthly token ceiling. Usage aggregates across agents and models until the cap is reached or your billing period resets.
OpenAI, Anthropic, Google Gemini, Mistral, xAI, DeepSeek, Kimi, and Llama-class endpoints appear in the same picker—routing chooses the model you selected per agent.
If you already have a provider contract, add keys under organization settings. When an active key exists, it takes precedence over platform-supplied credentials for that provider.
Some in-app surfaces still say FlexyBot when referring to hosted inference. The product brand is FlexyAgents—functionality is the same.
BYOK plans
BYOK plan slugs mirror hosted (free_byok, starter_byok, pro_byok, business_byok, enterprise_byok). Metering emphasizes messages and workspace limits rather than the hosted token-budget curve—check pricing for the exact entitlements.
Resolution logic loads organization keys from secure storage first. On BYOK plans, if no key exists for a provider, the API returns an explicit missing-key signal instead of silently using platform credentials.
FlexyAgents charges for the workspace, routing, knowledge, and channels. Token and tool usage billed by the vendor shows up on your existing cloud agreements.
BYOK plans are documented as message-based soft caps rather than the hosted token pool. That keeps billing predictable on our side while your provider handles token economics.
When a teammate picks a model but the org forgot the key, responses fail fast with a clear code—so you never assume an answer was grounded when inference never ran.
Decide
There is no hidden third mode—plans are either hosted-family or BYOK-family. You can change strategy when your contract renews; keys and agents migrate with the workspace.
You want a single vendor relationship, predictable FlexyAgents invoices, and the freedom to experiment across models without opening five cloud consoles on day one.
Legal or finance already standardized on OpenAI, Anthropic, or GCP enterprise deals and you need FlexyAgents as the orchestration layer—not another token middleman.
Provider coverage
The dropdowns in LLM API Keys mirror these vendors. Hosted plans expose the same catalog in agent settings—BYOK simply requires you to paste credentials before the first call.
10 vendors match Settings → LLM API Keys: OpenAI, Anthropic, Google Gemini, Azure OpenAI, Local / self-hosted, Mistral AI, xAI (Grok), DeepSeek, Kimi, Llama (Meta). Each has the model lists you see inside the dashboard.
Secrets are stored for the organization, never echoed back in API responses, and resolved only at inference time—matching the behavior documented in our LLM key resolution module.
Whether hosted or BYOK, each agent keeps its own default and fallback models. Changing plans does not erase agent configuration—only the funding path behind the scenes changes.
Wire models next to live data on Knowledge & RAG, browse SaaS connectors on Connectors, or see the rest of the surface on Features.
Next step
Start hosted to move fast, or choose BYOK if procurement already standardized on direct provider agreements. You can align the whole team on one path from the billing screen.