MCP Adr Analysis Server

What tosin2013/mcp-adr-analysis-server MCP server does

tosin2013/mcp-adr-analysis-server MCP server connects an MCP-compatible AI assistant with project code and Architectural Decision Records. It is intended for reviewing whether documented decisions still match implementation evidence, maintaining ADR context, and supporting architectural governance.

The project describes capabilities including technology-stack and architectural-pattern detection, drift detection, sensitive-content masking, decision memory, smart code linking, and deployment-readiness validation. Its session and tool-usage tracker stores project-local JSON snapshots and supports keyword-scored retrieval; it is not a graph database.

This server fits the Developer Tools category and is most relevant to teams that keep ADRs alongside active software projects. The optional ADR Aggregator integration can synchronize and share ADR context across teams.

How it works

The server communicates with an MCP client such as Claude Desktop, Cline, Cursor, or Windsurf. In the default CE-MCP mode, tools return orchestration directives that the host LLM uses to carry out analysis.

Tree-sitter provides incremental AST analysis for YAML and TypeScript and is used for semantic code understanding, function-signature extraction, and architectural-pattern discovery. Smart code linking combines keyword extraction with ripgrep searches to locate files related to ADRs and decisions. If native modules cannot be built, the server can run in a reduced mode without Tree-sitter code analysis.

Setup and configuration

The tosin2013/mcp-adr-analysis-server MCP server requires Node.js 20 or newer and npm 9 or newer. Install it globally with npm install -g mcp-adr-analysis-server, run it without a permanent installation through npx mcp-adr-analysis-server, or clone the repository and run npm install followed by npm run build.

A client configuration supplies the executable and the project location:

{
  "mcpServers": {
    "adr-analysis": {
      "command": "mcp-adr-analysis-server",
      "env": {
        "PROJECT_PATH": "/path/to/your/project"
      }
    }
  }
}

PROJECT_PATH identifies the project to analyze. The optional ADR Aggregator integration uses ADR_AGGREGATOR_API_KEY. Internet access is needed during installation for native module compilation; HTTP_PROXY and HTTPS_PROXY can be set for corporate proxy environments. An RHEL-specific installer is also provided.

Tools and capabilities

The listed MCP tools cover these operations:

• sync_to_aggregator: Send local ADRs to the optional platform.
• get_adr_context: Retrieve ADR context from that platform.
• get_staleness_report: Request ADR governance and health reports.
• get_adr_templates: Retrieve templates for specific domains.
• get_adr_diagrams: Obtain Mermaid diagrams for ADRs.
• validate_adr_compliance: Check an implementation against an ADR.
• get_knowledge_graph: Access a cross-repository knowledge graph.

The broader feature set also includes live drift checks against code and infrastructure evidence, security-oriented masking of secrets, PII, and other sensitive content, ADR generation and maintenance support, and deployment-readiness checks with hard test-failure blocking.

Limitations and notes

Native Tree-sitter analysis depends on successful installation and compilation of its modules. When those builds fail, the server remains usable in reduced mode but does not provide Tree-sitter code analysis. Installation requires network access unless the needed dependencies are already available.

The ADR Aggregator is optional rather than required for local analysis. The session and tool-usage tracker uses local JSON snapshots and keyword scoring, so it should not be treated as a general-purpose graph database. The repository lists an MIT license and requires a compatible MCP client.