ModelsAtlas
Explore
Atlas / MCP servers / robotmcp / ROS

ROSSAFE

mcp/robotmcp/ros-3
MCP server by robotmcp · mcp · mcp server · modelcontextprotocol · Audited 2026-09-24
⚑ Report / re-scanSource ↗

Connect AI models like Claude & GPT with robots using MCP and ROS.

Verdict
SAFE
Grade
B
Trust score
89 /100
Exposed tools
—
Transport
stdio · streamable-http
License
Apache-2.0
Stars
1,474
01

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

ROS-MCP-Server connects large language models (such as Claude, GPT, and Gemini) to robots, enabling bidirectional communication with no changes to existing robot source code.

Why ROS-MCP?

  • No robot source code changes → just add the rosbridge node to your existing ROS setup.
  • True two-way communication → LLMs can both control robots and observe everything happening on the Robot.
  • Full context → publish & subscribe to topics, call services & actions, set parameters, read sensor data, and monitor robot state in real time.
  • Deep ROS understanding → guides the LLM to discover available topics, services, actions, and their types (including custom ones) — enabling it to use them with the right syntax without manual configuration.
  • Works with any MCP client → built on the open MCP standard, supporting Claude Code, Codex CLI, Gemini CLI, Claude Desktop, ChatGPT, Cursor, and more.
  • Works across ROS versions → compatible across ROS 2 (Jazzy, Humble, and others) and ROS 1 distros.

🎥 Examples in Action

Read from source at commit 96438cba9b12OBSERVED · 2026-09-24
02

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.

claude-code (pypi)
claude mcp add ros-mcp -- None ros-mcp==3.1.2
03

Trust audit

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

LayerWhat it checksResult
L0Provenance & inventoryPASS
L1Static analysis of the codePASS
L2Instruction surface (what it tells the agent)PASS
L3Class-specific surfacePASS
L4Behavioural (sandbox)SKIPPED

What the source does

Filesystem
declared (1 observation(s))
Network
none-observed
Shell
none-observed
Dependencies
pinned
Secrets in source
none-found

Findings (21)

LOWInventory / provenance · inv.binary · CWE-1104
examples/3_limo_mobile_robot/isaac_sim/usd/limo_example.usd
limo_example.usd
Why it matters. a compiled or binary member cannot be reviewed from source
Fix. ship source, or explain the binary in the README
LOWNetwork egress · net.raw_ip · CWE-200, CWE-319
docs/install/from-source.md:46
Then point your AI client to `http://127.0.0.1:9000/mcp`. See the [HTTP transport](http-transport.md) page for client configuration details.
LOWNetwork egress · net.raw_ip · CWE-200, CWE-319
docs/install/http-transport.md:22
The server will start listening at `http://127.0.0.1:9000/mcp`.
LOWNetwork egress · net.raw_ip · CWE-200, CWE-319
docs/install/http-transport.md:42
"url": "http://127.0.0.1:9000/mcp"
LOWNetwork egress · net.raw_ip · CWE-200, CWE-319
examples/6_chatgpt/README.md:353
- Ensure the server is accessible at `http://127.0.0.1:9000/mcp`
LOWNetwork egress · net.raw_ip · CWE-200, CWE-319
examples/6_chatgpt/README.md:362
- Check if your domain is active: Visit ngrok dashboard at `http://127.0.0.1:4040/status`
LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94
ros_mcp/utils/websocket.py:141
image_bytes = base64.b64decode(data_b64)
LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94
ros_mcp/utils/websocket.py:155
image_bytes = base64.b64decode(data_b64)
LOWPrompt injection · prompt.credential_read · CWE-94, CWE-1427
examples/3_limo_mobile_robot/isaac_sim/README.md:118
Open PowerShell and configure the environment variables to enable ROS2 bridge:
Why it matters. asks the agent to read credentials
LOWPrompt injection · prompt.persistence · CWE-94, CWE-1427
examples/6_chatgpt/README.md:155
Add to `~/.zshrc` for persistence.
Why it matters. instructs the agent to persist itself in the user's environment
LOWSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357
docs/install/clients/chatgpt.md:14
curl -LsSf https://astral.sh/uv/install.sh | sh
LOWSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357
docs/install/clients/claude-code.md:10
curl -LsSf https://astral.sh/uv/install.sh | sh
LOWSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357
docs/install/clients/claude-desktop.md:13
curl -LsSf https://astral.sh/uv/install.sh | sh
LOWSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357
docs/install/clients/codex-cli.md:10
curl -LsSf https://astral.sh/uv/install.sh | sh
LOWSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357
docs/install/clients/cursor.md:10
curl -LsSf https://astral.sh/uv/install.sh | sh
LOWInventory / provenance · mcp.no_tools_extracted · CWE-1104
server.json
Why it matters. the tool list is enumerated at runtime by tools/list, not declared in source
Fix. the page says so rather than showing an empty table
INFOInventory / provenance · inv.oversize · CWE-1104
docs/images/MCP Demos Slide - 7to12s.gif
docs/images/MCP Demos Slide - 7to12s.gif
Why it matters. 6953500 bytes not read
INFOInventory / provenance · inv.oversize · CWE-1104
examples/3_limo_mobile_robot/images/limo.png
examples/3_limo_mobile_robot/images/limo.png
Why it matters. 2032689 bytes not read
INFOInventory / provenance · inv.oversize · CWE-1104
examples/3_limo_mobile_robot/images/limo_isaac_sim.png
examples/3_limo_mobile_robot/images/limo_isaac_sim.png
Why it matters. 1324096 bytes not read
INFOInventory / provenance · inv.oversize · CWE-1104
examples/3_limo_mobile_robot/images/limo_isaac_sim_simple_movement.gif
examples/3_limo_mobile_robot/images/limo_isaac_sim_simple_movement.gif
Why it matters. 17122573 bytes not read
INFOInventory / provenance · inv.oversize · CWE-1104
examples/3_limo_mobile_robot/images/limo_real_simple_movement.gif
examples/3_limo_mobile_robot/images/limo_real_simple_movement.gif
Why it matters. 49154186 bytes not read

Gates applied: no_behavioural_pass.

Audited 2026-09-24 · audit v0.4.1 · source sha 96438cba9b12full audit observations/trust-audit/mcp-server/robotmcp__ros-3.json · Report an issue / request a re-scan
04

Audit history

Every audit this server has had. A grade with a past is a grade somebody is still checking.

DateSourceVerdictGradeScoreChange
2026-09-2496438cba9b12SAFEB89first audit
05

Questions

What is the ROS MCP server?

Connect AI models like Claude & GPT with robots using MCP and ROS.

Is ROS safe to connect to an agent?

The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.

What credentials does ROS need?

No credential environment variables were found in its source, so it appears to need none.

How does ROS run?

It speaks stdio and streamable-http, so it runs as a local process your client starts. It is published on PyPI as ros-mcp.

How current is this page?

The grade is for one exact copy of the source (96438cba9b12), read on 2026-09-24. The repository is watched and re-audited when it changes.

Advertisement