Multi-Workspace MCP Server

Multi-Workspace MCP Server

Enables AI coding agents to access and manage multiple configured workspace repositories from a single session, eliminating context loss when working across companion repos.

Category
Visit Server

README

Multi-Workspace MCP Server

Enables coding agents to access configured repository workspaces from a single session, especially companion repos outside the current project workspace.

Problem Solved

AI coding agents are normally confined to a single workspace. When developing cross-platform features (e.g., FlowCoordinator on both Android/Kotlin and iOS/Swift), the agent loses all context when switching platforms. This MCP server gives the agent simultaneous access to both repositories, enabling:

  • Port features from Kotlin to Swift (or vice versa) without context loss
  • Read iOS code to learn conventions while working in Android Studio
  • Generate Swift files directly in the iOS repo
  • Search and inspect another repo without leaving the current agent session

Tools Provided

read_crossproject

Read a file from a configured workspace root — usually a companion repo outside the current project workspace. Prefer the IDE's built-in read/navigation tools for files in the current workspace. Returns up to 500 lines by default with line numbers and metadata (totalLines, startLine, endLine, truncated). Use offset/limit to paginate.

read_crossproject(workspace: "ios", path: "Modules/Messaging/Sources/MessagingCoordinator/StreamState.swift")
read_crossproject(workspace: "ios", path: "Modules/.../LargeFile.swift", offset: 100, limit: 50)

write_crossproject

Create or overwrite a file in a configured workspace root — typically another repo exposed through this MCP, not the current project. Prefer the IDE's built-in editing/refactoring tools for current-workspace files. If a writeAllowlist is configured, only matching paths are writable.

write_crossproject(
  workspace: "ios",
  path: "Modules/Messaging/Sources/MessagingCoordinator/v2/KeyReducer.swift",
  content: "// Swift code here"
)

list_crossproject

List files and directories in a configured workspace root, usually an external or companion repo rather than the current project workspace. Returns metadata (type, size, modified date) and supports recursive tree listing.

list_crossproject(workspace: "ios", path: "Modules/Messaging/Sources")
list_crossproject(workspace: "ios", path: "Modules", recursive: true, maxDepth: 2)

search_crossproject

Search text/regex in files inside a configured workspace root, usually a companion repo outside the current project workspace. Prefer the IDE's built-in code navigation/LSP tools in the current workspace when they apply. Supports output modes (content, files, count), context lines, path scoping (directory or file), and pagination.

search_crossproject(workspace: "ios", pattern: "FlowCoordinator", glob: "**/*.swift")
search_crossproject(workspace: "ios", pattern: "class.*Coordinator", path: "Modules/Messaging", outputMode: "files")

edit_crossproject

Apply search-and-replace edits to one or more files in a configured workspace root, typically for companion repos outside the current project workspace. Prefer the IDE's built-in refactoring/editing tools for typed changes in the current workspace. Supports regex with backreferences.

// Single file
edit_crossproject(
  workspace: "ios",
  paths: "Modules/.../MyFile.swift",
  oldString: "func oldName()",
  newString: "func newName()",
)

// Multiple files (same replacement applied to all)
edit_crossproject(
  workspace: "ios",
  paths: ["FileA.swift", "FileB.swift", "FileC.swift"],
  oldString: "oldValue",
  newString: "newValue",
  replaceAll: true,
)

// Regex with backreferences
edit_crossproject(
  workspace: "ios",
  paths: "Modules/.../MyFile.swift",
  oldString: "func (\\w+)\\(param: String\\)",
  newString: "func $1(param: Int)",
  useRegex: true,
)

run_crossproject

Run a shell command in a configured workspace root. Supports pipes, redirects, and chained commands. Returns stdout, stderr, and exit code. Use for build tools, codegen, git operations, or any command that needs to run in the workspace directory.

run_crossproject(workspace: "ios", command: "git status")
run_crossproject(workspace: "ios", command: "fastlane ios codegen_messaging")
run_crossproject(workspace: "android", command: "./gradlew :messaging:impl:test")
run_crossproject(workspace: "ios", command: "git log --oneline | head -5", timeout: 10000)

Configuration

Create a workspace-config.json in the repo root (see workspace-config.example.json):

{
  "workspaces": {
    "ios": {
      "root": "$HOME/git/zillow/ZillowMap",
      "name": "iOS (ZillowMap)"
    }
  }
}

By default, the agent can write to any file and run any command within the workspace root. To restrict either, add optional allowlists:

{
  "workspaces": {
    "ios": {
      "root": "$HOME/git/zillow/ZillowMap",
      "name": "iOS (ZillowMap)",
      "writeAllowlist": ["Modules/**/*.swift", "Tests/**/*.swift"],
      "runAllowlist": ["git", "fastlane", "xcodebuild", "swift"]
    }
  }
}
  • writeAllowlist (optional): Glob patterns. If omitted, all writes allowed. Controls write_crossproject and edit_crossproject.
  • runAllowlist (optional): Command prefixes. If omitted, all commands allowed. Controls run_crossproject. When configured, the first word of the command must match one of the prefixes (e.g., "git status" matches "git").

Supports $HOME and ~ expansion in root paths. There are no hardcoded defaults — all workspaces must be defined in this file. If missing, the server starts with zero workspaces and logs instructions.

Installation

npm install -g @exaudeus/workspace-mcp

Firebender Registration

Add to ~/.firebender/firebender.json:

{
  "mcpServers": {
    "workspace": {
      "command": "npx",
      "args": ["-y", "@exaudeus/workspace-mcp"]
    }
  }
}

Alternatively, if installed globally:

{
  "mcpServers": {
    "workspace": {
      "command": "workspace-mcp"
    }
  }
}

Usage Pattern

  1. Agent reads Android Kotlin file: read_file("libraries/illuminate/flow-coordinator/v2/KeyReducer.kt")
  2. Agent reads iOS conventions: read_crossproject("ios", "Modules/Messaging/Sources/MessagingCoordinator/SubjectCoordinator.swift")
  3. Agent references Rosetta mapping: (read .firebender/rosetta-kotlin-swift.md in Android workspace)
  4. Agent generates Swift equivalent
  5. Agent writes: write_crossproject("ios", "Modules/.../KeyReducer.swift", content)

All in one session, no context loss.

Companion Projects

  • Memory MCP: memory-mcp — persistent codebase knowledge for AI agents (separate repo)
  • Rosetta rules: .firebender/rosetta-kotlin-swift.md in Android workspace — idiom mapping reference

Development

# Install dependencies
npm install

# Build
npm run build

# Run in dev mode
npm run dev

# Run tests
npm test

Safety Features

Stateless Safety Model

The MCP uses stateless validation instead of session tracking - no brittle state to manage:

Edit Safety:

  • Verifies old string exists before editing (reads file fresh every time)
  • Enforces uniqueness (unless replaceAll=true)
  • Fails fast with clear errors if string not found or not unique
  • No need to "read first" - the verification IS the safety check

Write Safety:

  • Warns when overwriting existing files (logged to stderr)
  • Agent can still overwrite if intentional (not blocked)
  • Optional writeAllowlist restricts writes to matching paths (omit to allow all)

Why Stateless?

  • No session state = no brittleness across restarts/reconnects
  • Always reads fresh from disk (catches external changes)
  • Works across multiple agents/sessions
  • Self-documenting (failures explain what's wrong)

Other Safety Features

  • Write allowlist (optional): When configured, only allows writes to paths matching the allowlist patterns
  • Path validation: Resolved paths are verified to stay within workspace root (prevents directory traversal)
  • Shell injection prevention: All external commands use execFile with argument arrays (no shell interpolation)
  • Audit logging: All write/edit operations logged to stderr

Future Enhancements

  • Git operations (git_crossproject)
  • Additional workspaces (backend repos, etc.)
  • Auto-context loading (inject .firebender/platform-context.md on first access)

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured