AI & MCP

MCP Overview

How SPECTRALE uses the Model Context Protocol to give compatible AI clients structured architecture context.

Give your AI agents the bigger picture.

AI coding agents are increasingly capable of understanding code, editing files, and assisting with complex development tasks.

But a repository is only one part of a software system.

An agent working inside a codebase may not know which external services the application depends on, how other components relate to it, or what infrastructure exists beyond the files currently available.

SPECTRALE helps bridge that context gap.

Through the Model Context Protocol (MCP), compatible AI clients can request structured information about your software architecture.

What is MCP?

The Model Context Protocol is an open protocol for connecting AI applications to external tools and contextual information.

An MCP server exposes a defined set of capabilities that compatible clients can discover and invoke.

SPECTRALE provides an MCP interface for accessing architectural information.

How it works

text

AI DEVELOPMENT CLIENT
        |
        |  MCP
        v
SPECTRALE MCP SERVER
        |
        |  Authorization
        v
PROJECT ARCHITECTURE
        |
        +-- Components
        +-- Dependencies
        +-- Discovery status
        +-- External boundaries
        +-- Integration metadata

The AI client requests information through MCP.

SPECTRALE checks the applicable permissions and retrieves information from the authorized architecture context.

The client receives structured results that it can use within its own workflow.

Why architectural context matters

Consider a developer asking an AI agent:

"Which services could be affected if I change this database?"

Without architectural context, the agent may need to infer dependencies from the currently accessible repository.

With SPECTRALE, a compatible agent can request the project's known components and dependencies.

That doesn't guarantee a complete impact analysis. It provides an additional source of structured evidence for the agent to consider.

Designed for different AI clients

SPECTRALE's MCP interface is based on an open protocol rather than a proprietary agent integration.

The goal is to make architecture context available across compatible MCP clients, without requiring a separate SPECTRALE-specific integration for every AI development environment.

Client support and authentication compatibility depend on the client's MCP implementation.

Read-only by design

SPECTRALE's current MCP toolset focuses on retrieving architectural information.

The tools are designed to inspect project context, not modify infrastructure.

MCP access does not inherently grant permission to:

Deploy applications.
Modify databases.
Change cloud infrastructure.
Create third-party connections.
Execute arbitrary commands.

This separation allows architectural context to be used without turning the MCP server into a general-purpose infrastructure control interface.

Deployment handoff is a separate Architect workflow. It does not turn these read-only context tools into infrastructure controls. See Deploy on Agent and Deployment Security.

Connected agents that use this interface are coordinated in Fleet. Questions about the architecture those tools return can be explored in Guardian.