The MCP Manager is the control plane for Britive's MCP integration: register backend MCP servers and their tools, define credential providers, and set the policies that govern which identities—human or AI agent—can invoke which tools. The MCP Gateway is the data plane: gateway agents proxy client tool requests to those backend servers and check out credentials for them at request time.
MCP Servers and Tools
An MCP server is a logical container that groups a collection of tools. Each tool is an independent capability exposed by the server that can be invoked by an MCP client, for example, querying a database, calling an external API, or retrieving a secret.
Tools are the fundamental unit of access within the MCP platform: they are accessed by clients, protected by authorization policies, and associated with one or more credential providers. Platform policies and associated permissions determine who can create, modify, and manage MCP servers and their tool definitions.
For more information about creating and managing an MCP server, see MCP servers.
Credential Providers
Credential providers define how credentials are resolved during tool access. Each provider is attached to a tool and implements one of the following resolution strategies:
Application profile checkout
Resource profile checkout
Secret retrieval from the vault
When a tool is accessed, the MCP Manager resolves the credential providers associated with the tool, performs the checkout against the appropriate backend, and returns the resolved credentials asynchronously to the requester.
For more information about creating and managing credential providers, see Credential Providers.
Gateways
The gateway authenticates to the platform to read the MCP server catalog and publish tools. For more information, see Gateways.
Policy-Based Authorization and Access Governance
A policy attached to an MCP server defines the following:
Which identities can access the server
Which tools are available to those identities
Which credential providers may be used for each tool
These give precise control over the credentials returned during tool access.
On a tool access request, the platform:
Resolves all applicable policies
Verifies the requesting identity is authorized
Evaluates policy conditions (such as IP restrictions and time-based access windows)
Returns only the credential providers explicitly permitted by matching policies.
There is no implicit access: every credential a tool can return must be explicitly authorized through a policy. Policies support draft, enabled, and disabled states, and policy ordering determines deterministic precedence when multiple policies grant the same tool.
For more information about creating and managing policies, see MCP Servers -> Policies.
Britive MCP Gateway
The Britive MCP Gateway is positioned between the AI initiating tool calls and the MCP servers that execute them. Every request is routed through the gateway, which verifies whether the requesting identity is authorized, issues the backend a time-limited credential, and records the resulting activity.
For more information about creating and managing an MCP gateway, see Britive MCP Gateway.