When to use direct access

Use direct access when ToolPlexer is an implementation detail of your product. Your backend owns one workspace and calls ToolPlexer server-to-server.

Use direct access when

• ToolPlexer powers enrichment, background jobs, or internal agents.
• You want one server-side key and no user credential flow.
• You bill users through your own subscription, credits, or feature limits.
• You want one wallet and one usage surface for the product.

Two patterns

Full discover / inspect / run

Let your backend or internal agent select capabilities dynamically. Use this when the required provider is not known ahead of time:

discover → inspect → run

Curated endpoints only

Pre-select a small set of endpoints and expose them through your own functions. This is best when your interface is fixed and pricing must be predictable.

Use another model when

• users should own their own workspace and spend — use user-owned access
• you need your own billing or endpoint allowlist — use a proxy

Examples

• A CRM enriches new leads in a background job.
• A support product adds a “research this ticket” action.
• A nightly analytics pipeline refreshes a curated dataset.