NewApp Monitoring is liveSign up free
AgentbeamBeta
AI agent vulnerability · OpenClaw

GHSA-qh2f-99mv-mrcf: OpenClaw: Bundle MCP loopback could miss its exec denylist on session spawn

mediumCVSS 4.0 6.9

OpenClaw's bundled MCP loopback skipped its exec denylist when a caller spawned a session. The spawned session could then run commands that the loopback entry point was supposed to block. Versions before 2026.5.12 are affected. Update the openclaw package to 2026.5.12 and limit loopback access to trusted operators.

Published Jul 2, 2026 · Analysis Oct 25, 2026

What an attacker can do

The bundled MCP loopback is an entry point that carries an exec denylist, a list of commands it should not allow. On the session-spawn path in affected versions, that denylist was not applied. A caller who could reach this path started a session with wider command reach than the loopback should give. The weaknesses are listed as OS command injection (CWE-78) and improper access control (CWE-284). The advisory does not give the denylist entries or the tool names.

The CVSS 4.0 vector is network, low complexity, no privileges, no user interaction, with low impact on confidentiality, integrity and availability. The attacker must be able to reach the bundled MCP loopback. Commands run with the Gateway's rights. A prompt-injected agent that uses MCP tools could trigger the spawn path too.

The advisory leaves OpenClaw's trusted-operator model unchanged. The risk is in setups where lower-trust users or content can reach this path.

Who is affected

Affected: npm openclaw before 2026.5.12. You are exposed if the bundled MCP loopback is enabled and reachable by callers you do not fully trust.

npm ls -g openclaw

Drop -g for a project install. Check your Gateway configuration for the bundled MCP loopback.

openclaw

Ecosystem
npm
Affected
< 2026.5.12
Fixed in
2026.5.12
Check
npm ls openclaw

How to fix it

Update to openclaw 2026.5.12, the first stable patched version.

npm install -g [email protected]

Until you can update, restrict bundled MCP loopback access to trusted operators. The advisory also suggests narrow channel and tool allowlists, not sharing one Gateway between mutually untrusted users, and disabling the feature if you do not need it.

How to tell if you were hit

The advisory gives no log line. Review sessions spawned through the bundled MCP loopback and the commands they ran. Compare those commands with your exec denylist, and look for ones it should have blocked. Check the process tree under the Gateway for unexpected shells or tools. Beam records agent shell commands and MCP tool calls, and its rules flag risky ones, which helps with this review.

Original advisory

From the GitHub Advisory Database (CC BY 4.0).

Summary

Bundle MCP loopback could miss its exec denylist on session spawn. In affected versions, a caller that can reach the affected bundled MCP session-spawn path could bypass the denylist that was intended for that loopback MCP entry point.

This advisory is scoped to the named feature and configuration. It does not change OpenClaw's trusted-operator model: authenticated Gateway operators, installed plugins, and intentional local execution surfaces remain trusted unless a separate policy, approval, allowlist, sandbox, or auth boundary is crossed.

Impact

When the affected feature is enabled and reachable, this could start a session with broader command reach than that MCP path should provide. Practical impact depends on the operator's configuration and whether lower-trust input can reach that path.

Patched Versions

The first stable patched version is 2026.5.12.

Mitigations

restrict bundled MCP loopback access to trusted operators until patched. As general hardening, keep channel and tool allowlists narrow, avoid sharing one Gateway between mutually untrusted users, and disable the affected feature when it is not needed.

Credited to cantinagen, Ellahinator.

Questions

Which OpenClaw version fixes the bundled MCP loopback denylist bypass?

2026.5.12. All earlier versions are affected.

Is this a remote code execution bug?

The CVSS vector is network with no privileges, but the advisory describes broader command reach on a spawned session, not a direct code execution flaw. Exposure depends on who can reach the loopback.

Is there a workaround?

Restrict bundled MCP loopback access to trusted operators, keep allowlists narrow, and disable the feature if you do not need it.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology

More OpenClaw vulnerabilities

All OpenClaw entries