Glossary

Terms used across this documentation. For how the main ideas relate, see Core Concepts.

LockMCP terms

Workspace Your organisation's account. Contains connections, tools, members, roles and audit history. Isolated from every other workspace.

Connection One API you're exposing to AI clients. Holds a base URL, credentials, an identity model, and its own MCP endpoint. A workspace can have many, each configured independently.

Identity model How your users are recognised — Identity Brokering, Token Exchange or Tenant Delegated. Chosen per connection.

Tool One action an AI client can take, usually one API operation. Generated from an OpenAPI spec, refined in Tool Studio, published in the Tool Catalog.

Draft / Published A tool's state. Drafts are never served to AI clients; publishing makes a tool available — subject to role access.

Role A named bundle of access assigned to members. Carries both dashboard permissions and tool access.

Permission What someone can do in the dashboard (e.g. tools.publish, members.invite). Distinct from tool access.

Tool access Which tools a role's members may call. Separate from publishing — both are required for a tool to be visible.

Member A person who can use your connections.

MCP endpoint The URL an AI client connects to. Each connection has its own.

Provisioning Wizard Step 4. Sets up a connection's identity resources — or, for Tenant Delegated, runs certification against your endpoints.

Certification The seven-check suite run against Tenant Delegated endpoints during provisioning. All must pass. See Tenant Delegated Troubleshooting.

Domain verification Wizard Step 2. Proves you control your domain, via DNS TXT record or a well-known file. Gates provisioning for every identity model.

Enum injection Constraining a tool parameter at runtime to the resource IDs the current user may access. Available on all three identity models. See Enum Injection.

Inject key The key name linking an x-mcp-inject marker in your spec to values in the bootstrap payload.

Bootstrap payload The role-and-scoped-ID data enabling enum injection, returned in your validate response.

Identity model terms

Identity Brokering Users sign in through LockMCP directly. No tenant-side setup. Works with desktop AI clients.

Token Exchange You register your existing identity provider, which LockMCP federates. Does not support interactive browser clients.

Tenant Delegated Your application owns the user directory and login page. Your backend implements a contract and is consulted on tool calls.

Federation Registering your identity provider so LockMCP trusts sign-ins from it.

Role mapping Mapping your application's role names onto LockMCP roles (Tenant Delegated). Unmapped names are ignored — they don't fall back to a default.

Standards and protocols

MCP (Model Context Protocol) The open protocol AI clients use to discover and call tools.

OpenAPI The specification format describing your API. LockMCP parses it to generate tools. See Writing OpenAPI Specs.

OAuth 2.0 The authorisation framework behind the browser sign-in flow.

OIDC (OpenID Connect) The identity layer on OAuth 2.0. Token Exchange requires an OIDC-compliant provider.

JWT (JSON Web Token) The signed token format carrying a user's identity and claims.

JWKS (JSON Web Key Set) The public keys used to verify tokens from an identity provider. Its URL is one of two fields Token Exchange requires.

Issuer The iss claim identifying who minted a token. Compared as an exact string — a trailing slash makes it a different value.

Claim A field inside a JWT (e.g. sub, exp, iss).

sub (subject) The claim identifying the user. LockMCP resolves members from it.

RS256 / ES256 Signature algorithms. RS256 signs hub-issued tokens for Identity Brokering and Token Exchange; ES256 is used for Tenant Delegated.

HMAC (HMAC-SHA256) The signature scheme on Tenant Delegated server-to-server requests, proving they came from LockMCP. See HMAC Signing.

PKCE An OAuth extension protecting the authorization-code exchange. Your login page must tolerate its parameters.

RFC 8693 The OAuth Token Exchange standard the Token Exchange model is modelled on.

DNS TXT record A text record on your domain, one of the two domain-verification methods.

Well-known file A file served at a specific path on your domain — the other verification method.

SSRF (Server-Side Request Forgery) An attack making a server request internal addresses. LockMCP validates outbound tool requests against it.

AES-256-GCM The authenticated encryption protecting stored credentials. See Security.

Where to go next