Give every customer their own governed AI integration.
Expose one shared API project to many downstream organizations, each with isolated credentials, users, access, and connect flows.
The B2B2C reality
SaaS platforms rarely need a completely different MCP server per customer. They need one product API, one evolving spec, and a way to let each customer consume it with the right credentials and permissions.
- The API surface is shared. Every customer uses the same product capabilities, so duplicating infrastructure per customer creates drag.
- The credentials are not shared. Each customer still needs separate access to its own backend resources.
- Onboarding must be self-serve. Prospects and customers should not need a human provisioning loop to start testing.
- Access needs to map to organizations. Teams, users, and AI clients should inherit the right boundaries automatically.
How Frege models it
Frege keeps your API projects at the tenant level and layers downstream organizations beneath them. The spec, docs, and versions are shared. Credentials, users, OAuth clients, and access grants stay organization-specific.
The connect flow is the product
Instead of sending customers a setup document full of manual steps, you share a connect page. The page can show MCP endpoint details, client-specific configuration snippets, and a path to create or retrieve the right credentials.
When a customer authenticates through the connect flow, Frege can auto-create the organization, issue a project-scoped API key, and display a ready-to-use MCP configuration immediately. This shortens the path from interest to working integration.
Branded onboarding
The connect page lives on a shareable project URL and explains how to connect Claude Desktop, Cursor, and generic MCP clients. That lets product teams treat MCP access like a first-class feature instead of a support-managed exception.
Guest access for evaluation
Frege also supports time-limited guest keys so prospects can try the integration before full signup. That is especially useful for enterprise evaluation or partner enablement workflows.
Per-customer credentials without per-client glue
Customers usually already have their own auth model for your API: bearer tokens, API keys, refresh tokens, custom headers, HMAC signing, or some combination. Frege stores those credentials per organization and applies them through the proxy layer at request time.
That means a customer configures access once and every AI client connected through Frege inherits the right backend auth automatically.
Why this works well for SaaS teams
- One spec, many customers: maintain one versioned MCP project instead of duplicating tool definitions across tenants.
- Organization-safe credentials: preserve per-customer isolation while reusing the same product surface.
- Self-serve onboarding: connect pages and guest sessions reduce support load.
- Shared rollout controls: publish new versions once, then let downstream customers migrate with explicit version pinning.
Ship MCP like a platform feature.
Keep one shared API surface, give every customer isolated access, and turn onboarding into a connect flow instead of a support queue.