Skip to content

PATTERN Cited by 1 source

MCP as context bridge

Definition

MCP as context bridge is a pattern where Model Context Protocol is used to compose multiple, previously disconnected context sources into a single agent session so that a model can reason across them jointly. The protocol eliminates the need for bespoke integrations per source system — an agent connects to an MCP server that already indexes the relevant content.

Mechanism

Agent session
  ├── [MCP tool call] ──► MCP Server A (indexed docs, threat models, design reviews)
  ├── [direct input] ──► Code change / PR content
  └── Model reasons across BOTH → identifies gaps, inconsistencies, alignments

The key insight: the value isn't just retrieving a document — it's giving the model access to both sides of a relationship (e.g., requirement + implementation) simultaneously so it can reason about their alignment.

Canonical instance

Dropbox's security review agent uses the Dash MCP server to retrieve threat models during code review. The model then compares documented security requirements against the proposed code change — something neither document alone could reveal (Source: sources/2026-06-12-dropbox-mcp-dash-design-to-code-security).

"MCP composes multiple context sources into a single agent session. The model reasons across them to identify gaps between security requirements and implementation."

Why MCP specifically

  • Single integration. The agent connects to one MCP server rather than building connectors for every documentation system.
  • Already-indexed content. Dash's MCP server exposes the same search index that powers user-facing Dash search — years of threat models are searchable without additional ingestion.
  • Protocol-level interop. Any MCP-compatible agent can use the same server; the pattern isn't locked to one client.

Relationship to other patterns

Seen in

Last updated · 542 distilled / 1,571 read