Skip to main content

Bring Your Own AI

Bring Your Own AI lets a workspace route Woes support-agent model calls through the customer’s own model-provider account. It is a dedicated plan, not a normal feature toggle.

Availability

Bring Your Own AI appears in Settings only after the workspace is on the Bring Your Own AI plan. If the workspace is not on that plan, Woes shows a plan-required state and links admins back to Billing.

Supported Providers

Woes supports a curated provider list in the app, currently including Anthropic, OpenAI, and Google where enabled. The available providers and models may change as Woes updates its server-side routing and reliability controls. Each provider entry can have one stored key. A workspace can connect more than one provider, but only one provider is active for support-agent routing at a time.

Frontier Model Selection

Woes exposes a curated model picker for the active provider. The picker prevents unsupported free-form model strings from being routed at runtime. If a stored or submitted model id is empty, stale, or unsupported, Woes falls back to the provider default configured by the platform.

Setup Flow

1

Switch to the plan

Open Settings → Billing and switch to Bring Your Own AI.
2

Open Bring Your Own AI

After the plan is active, open Settings → Bring Your Own AI.
3

Choose a provider

Select one of the supported providers shown in the app.
4

Paste the provider key

Paste the provider API key from the provider console. Woes never shows the full key again after it is saved.
5

Pick a model

Choose one of the supported frontier models. If no model is selected, Woes uses the provider default.
6

Verify and activate

Woes verifies the key and can activate the first connected provider so the support agent has a provider to use.

BYO AI Onboarding

When a workspace is on the Bring Your Own AI plan and has not connected a provider, Woes can guide admins through a focused setup wizard. The wizard:
  • Explains the plan and provider routing behavior.
  • Walks through provider-key setup.
  • Verifies connected keys.
  • Activates the first connected provider by default.
  • Lets admins skip setup for the current session.
Until a provider is active, Woes uses the platform AI routing path.

Runtime Behavior

When an active BYO AI provider exists, the support agent uses:
  • The customer’s active provider.
  • The customer’s decrypted provider key on the server.
  • The selected supported model, or the provider default.
  • The same Woes retrieval, grounding, confidence, clarification, and handoff rules.
BYO AI changes the model execution account. It does not change tenant boundaries, source retrieval, redaction, confidence rules, or what customers are allowed to see.

Key Security

Provider keys are private credentials. Woes stores provider keys server-side and returns only a redacted key hint such as the last four characters. Customer-facing widgets, REST API responses, operator debug traces, and public docs must never expose the full key. Do not paste provider keys into:
  • Agent prompts.
  • Source documentation.
  • Macros.
  • Customer replies.
  • Public docs or screenshots.
  • Browser-side widget configuration.

Operations

Admins can:
  • Connect a provider key.
  • Test a provider key.
  • Activate a provider.
  • Pause BYO routing by returning to platform AI.
  • Remove a provider key.
  • Update the model for a connected provider.
If a provider key fails verification, Woes can save the status as failed and show an operator-safe error. The full key remains hidden. Failed verification does not activate the provider for support-agent routing. Pausing BYO AI returns the workspace to platform-managed routing without changing agent grounding, source attachments, confidence, or handoff behavior.

Billing Notes

Bring Your Own AI is billed by Woes as a plan. Model usage itself is billed by the customer’s provider account according to that provider’s pricing. Review Billing for plan status, subscription management, and workspace capacity.
Last modified on July 3, 2026