Previously “StoryMaps”

MCP Server

A cross-repo MCP server
for your entire codebase.

Give Cursor, Claude Code, or VS Code structural context that spans every repository in your organization. Not just the repo you have open — all of them.

Single-repo context tools miss the full picture.

Your IDE's built-in AI reads the files you have open — maybe the whole repo. But production systems span dozens of repositories. ArcLume's codebase MCP server bridges the gap.

Cursor's @codebase

Indexes the repo you have open. Great for single-repo questions. Blind to the 15 other services your endpoint talks to.

Claude Code's tools

Reads files and runs commands in the current project. Powerful locally, but can't trace a Kafka topic consumer in a repo you haven't cloned.

ArcLume MCP

Queries a pre-built structural model spanning every connected repository. Cross-repo dependencies, call graphs, and interface points — all resolved before your AI even asks.

Questions your IDE can finally answer

With ArcLume's MCP server connected, your AI assistant can answer structural questions that were previously impossible without manual code archaeology.

>

"What services will break if I rename this endpoint?"

ArcLume traces every consumer of the endpoint across all repos and returns the affected files, functions, and confidence levels.

>

"Show me every repo that imports from shared-types"

Returns a list of every repository and file that depends on the shared package, with the specific symbols they import.

>

"What's the call chain from the /checkout endpoint to the payment provider?"

Traces the full execution path from the HTTP handler through service layers, queues, and external API calls — across repository boundaries.

>

"Which repos handle the user.deleted event?"

Finds every consumer of the event across all services, including the handler function, file path, and whether the match is confirmed or heuristic.

Set up in three steps

1

Connect repos via the ArcLume dashboard

Install the GitHub App and select the repositories you want in your structural model. ArcLume indexes them automatically and re-indexes on every push.

2

Add ArcLume as an MCP server in your IDE

Copy your MCP server URL from the ArcLume dashboard, then add it to your IDE's MCP configuration.

Cursor — .cursor/mcp.json

{
  "mcpServers": {
    "arclume": {
      "url": "https://mcp.arclume.dev/sse?token=YOUR_TOKEN"
    }
  }
}
3

Start querying

Your AI assistant now has access to ArcLume's tools: search the codebase, trace dependencies, list stories, and query the structural model. Ask any cross-repo question in natural language.

Works everywhere MCP does

ArcLume exposes a standard MCP server over SSE. Any editor or tool that supports the Model Context Protocol gets full access.

Cursor

Native MCP support

Claude Code

Native MCP support

VS Code

Via MCP extension

Your AI deserves cross-repo context.

Set up ArcLume's MCP server in minutes and give your IDE AI the structural context it needs to answer real architecture questions.

Connect your IDE