About ROS 2 Workspace Inspector
ROS 2 Workspace Inspector is a local, read-only MCP server for examining ROS 2 workspaces without installing ROS 2. It provides bounded static analysis for robotics developers and keeps file access within a configured workspace root.
Key features
- 01Discover workspace packages and metadata
- 02Inspect package structure and interfaces
- 03Analyze dependencies and cycles
- 04Review launch files and robot descriptions
- 05Run bounded workspace diagnosis
Use cases
- 01Map packages in a ROS 2 workspace
- 02Check dependency order and cycles
- 03Review launch files without executing them
- 04Inspect URDF and Xacro descriptions
MCP support
It exposes seven static-analysis tools, two read-only resources, and a guided review prompt to MCP clients over stdio. The server reads only within the configured root and does not execute workspace files.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEROS 2 Workspace Inspector MCP
ROS 2 Workspace Inspector MCP is a local-first, read-only static-analysis MCP server that gives Codex, Claude Code, Cursor, and other MCP clients structured information about ROS 2 workspaces. Seven static-analysis tools cover discovery, package structure, dependencies, interfaces, launch files, robot descriptions, and bounded workspace diagnosis without requiring ROS 2. It also provides two compact read-only resources and one guided review prompt.
7 analysis tools · 2 read-only resource forms · 1 guided review prompt
The project inspects files only. It does not control robots, run ROS nodes, build workspaces,
execute launch or Xacro files, execute setup.py, publish topics, call services or actions, or
modify user files. All access stays inside the configured root sandbox. Only stdio is supported.
Tools
scan_workspace— Discover ROS 2 packages, workspace layout, build types, and basic manifest metadata.inspect_package— Inspect a package's manifest, build configuration, executables, launch files, interfaces, robot descriptions, and tests.analyze_dependencies— Analyze package dependencies, internal dependency edges, topological order, and dependency cycles.inspect_interfaces— Parse and inspect ROS 2 Msg, Srv, and Action interface definitions.analyze_launch_file— Statically analyze a Python, XML, or YAML ROS 2 launch file without executing it.inspect_robot_description— Inspect URDF structure and statically summarize Xacro files without expanding or executing them.diagnose_workspace— Run a consolidated, bounded diagnosis of the configured ROS 2 workspace.
Requirements and installation
- Python 3.10 or newer
uv
Clone the repository, enter it, and create the locked environment:
git clone https://github.com/doer0312/ros2-workspace-mcp.git
cd ros2-workspace-mcp
uv sync --locked
Run against a local workspace:
uv run ros2-workspace-mcp --root /absolute/path/to/your/ros2_ws
The equivalent module entry point is:
uv run python -m ros2_workspace_mcp --root /absolute/path/to/your/ros2_ws
The process communicates via stdout using the MCP protocol. Do not redirect ordinary logs to stdout. Invalid CLI input is reported on stderr with a non-zero exit status.
Client configuration
Codex TOML:
[mcp_servers.ros2-workspace-inspector]
command = "uv"
args = ["--directory", "/absolute/path/to/ros2-workspace-mcp", "run", "ros2-workspace-mcp", "--root", "/absolute/path/to/your/ros2_ws"]
Claude Desktop JSON:
{
"mcpServers": {
"ros2-workspace-inspector": {
"command": "uv",
"args": ["--directory", "/absolute/path/to/ros2-workspace-mcp", "run", "ros2-workspace-mcp", "--root", "/absolute/path/to/your/ros2_ws"]
}
}
}
Copyable versions are available under examples/client-configs/. Both examples run the checked-out
repository environment and do not assume a PyPI release.
Try the included static demo without installing ROS:
uv run ros2-workspace-mcp --root ./examples/demo_ws
Minimal request:
Scan the configured ROS 2 workspace and summarize its packages.
Guided review example:
Use review_ros2_workspace with depth="standard".
Available tools
The server exposes exactly seven tools:
scan_workspace— discover packages and basic metadata.inspect_package— inspect one package's manifest, build files, executables, and file layout.analyze_dependencies— parse declarations and build a workspace dependency graph.inspect_interfaces— parse one package's.msg,.srv, and.actionfiles.analyze_launch_file— statically analyze one Python, XML, or YAML launch file.inspect_robot_description— validate URDF or summarize unexpanded Xacro.diagnose_workspace— run a bounded, deduplicated workspace-wide diagnosis.
Example requests:
Use scan_workspace to inspect the configured ROS 2 workspace.
Inspect the package named demo_robot.
Analyze src/demo_robot/launch/demo.launch.py without running it.
Diagnose the configured workspace.
scan_workspace takes no arguments and scans only the workspace supplied with --root. A compact
result looks like this:
{
"layout": "colcon_workspace",
"package_count": 1,
"packages": [
{
"name": "demo_pkg",
"relative_path": "src/demo_pkg",
"build_type": "ament_python",
"build_type_source": "inferred",
"valid": true
}
],
"issues": []
}
The full result also includes server identity, canonical root and scan paths, package metadata,
valid/invalid counts, duplicate names, and structured issues. See docs/tools.md.
Resources and prompt
ros2-workspace://summaryis a fixed compact workspace summary.ros2-workspace://package/{package_name}is a package context template with strict name validation and ambiguity errors.review_ros2_workspace(focus="", depth="standard")guides quick, standard, or deep review using the seven tools without reading the workspace while the prompt is generated.
Resources are generated on read, have application/json MIME type, and never return raw project
files. The server intentionally does not expose an arbitrary file resource, subscriptions,
notifications, or filesystem watching. Client applications may display resources and prompts in
different ways. See docs/resources.md and
docs/prompts.md.
Security boundary
The server is bound to one validated, canonical workspace directory. Every file selected for
reading is resolved and checked with path containment rather than string prefixes. Directory
symlinks are never traversed; readable file symlinks must resolve inside the root. Text reads are
UTF-8, size-limited, and centralized. Python is parsed with AST, XML with ElementTree, and YAML
with safe_load; setup, launch, Xacro, CMake, and process declarations are never executed. See
docs/security.md for the complete boundary.
Development
uv run ruff check .
uv run ruff format --check .
uv run pytest
Further documentation:
License
Licensed under the Apache License 2.0. See LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.