Security
Overview
Lightcast MCP currently implements service-to-service authentication, schema validation, and infrastructure-level security controls for hosted deployments.
This page describes the current security model, including authentication, authorization boundaries, deployment controls, logging behavior, and known limitations.
Customer-facing authentication, authorization, tenant isolation, and entitlement management capabilities are not currently available in the documented product surface. Documentation for these features will be published soon.
Current Authentication Model
Lightcast MCP uses service-to-service authentication to communicate with downstream Lightcast services.
When the MCP server receives a request, it uses configured Lightcast client credentials to obtain access tokens through the OAuth 2.0 Client Credentials flow. These tokens are then used to access the required downstream services.
| Downstream Scope | Used For |
|---|---|
postings:us | Job postings data |
profiles:us | Profiles data |
agnitio | Core LMI
|
To improve performance and reduce authentication overhead:
- Access tokens are cached based on scope.
- Cached tokens are refreshed before they expire.
- If a downstream service returns an authentication error, the cached token can be invalidated and a new token requested automatically.
MCP Client Authentication
Lightcast MCP does not currently provide a documented client authentication mechanism for MCP consumers.
The following capabilities are not currently available:
- MCP OAuth authorization code flow.
- MCP bearer-token validation.
- API key validation at the MCP endpoint.
- User identity extraction.
- Session persistence.
- Customer-specific credential exchange.
Tool traces may include an MCP session identifier provided by the MCP context. This identifier is used for tracing purposes only and should not be considered an authentication or authorization mechanism.
Authorization and Entitlements
Lightcast MCP validates that requests use supported datasets, tools, dimensions, metrics, and schema definitions. The platform does not currently implement customer-specific authorization or entitlement enforcement at the MCP layer.
Access is currently governed by the following controls:
| Control | Description |
|---|---|
| Supported dataset and tool matrix | Only configured datasets and supported tools can be called successfully. |
| Tool schema validation | Unsupported dimensions, metrics, and input shapes are rejected. |
| Downstream credentials | Access to downstream services is controlled by the configured service credentials and OAuth scopes. |
Logging and Tracing
Lightcast MCP includes structured logging and tracing:
- HTTP requests
- Tool calls
- Authentication calls
- Downstream service calls
Downstream requests include a correlation identifier derived from the active trace context. Implemented safeguards include redaction of downstream access tokens in authentication response logs.
👉 Request-body redaction for all tool inputs and downstream payloads is not documented as a released capability.
Treat tool arguments and downstream response content as potentially sensitive when configuring logs.
Current Limitations
The following capabilities are not validated as released customer-facing security features:
- MCP endpoint authentication
- Customer-specific authorization
- Tenant isolation
- Package-level entitlement enforcement
- MCP-layer rate limiting
- CORS policy controls
- Standardized customer identity propagation to tool handlers
See Errors for information about structured error responses and error-handling behavior.
Updated 16 days ago