Latest release: v2026.7.1Download zip
Compatibility
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The plugin provides the Codex app-server harness, Codex model/provider catalog, media understanding, web search, migrations, and native Codex thread controls; high-impact capabilities such as command execution, thread management, diagnostics, and auth bridging are purpose-aligned for a Codex runtime integration.
Instruction Scope
Activation is limited to Codex agent harness use rather than startup, and sensitive actions such as diagnostics upload and computer-use installation include explicit commands or owner/user controls; advanced options can intentionally enable broader Codex permissions.
Install Mechanism
The package is the trusted official @openclaw/codex release, source-linked to openclaw/openclaw, with ordinary npm dependencies and no evidence of deceptive install behavior in the package metadata.
Credentials
The plugin may spawn or connect to Codex app-server, read configured Codex/OpenAI auth, use WebSocket transport, and route sandboxed filesystem/network/process requests; these accesses are substantial but coherent with running Codex-backed agent turns.
Persistence & Privilege
It persists Codex session bindings and can create an agent-scoped Codex home or optionally use the user's Codex home; migration paths can copy skills/config/auth only through migration workflows, with hooks archived for review rather than activated automatically.
Scan Findings in Context
[suspicious.dangerous_exec] expected: Static scan flagged child_process spawn sites; review shows they start the managed Codex app-server, resume Codex CLI sessions, or run OpenClaw sandbox-backed subprocess helpers, which is expected for this harness.
[VirusTotal clean] expected: VirusTotal telemetry reported no malicious or suspicious engine detections; this supports but does not by itself determine the benign verdict.
[SkillSpector none] expected: No SkillSpector findings were supplied, and artifact review did not identify hidden exfiltration, deceptive instructions, or purpose-mismatched persistence.
Assessment
Install this if you want OpenClaw to run Codex-backed agents. Review advanced settings before enabling user home sharing, danger-full-access/yolo permissions, native Codex plugins, computer-use auto-install, remote WebSocket app-server endpoints, or diagnostics upload, because those options can expose local files, credentials, or connected app actions to Codex workflows.dist/command-rpc-eLM_mtC7.js:424
Shell command execution detected (child_process).
dist/run-attempt-FUyOjGCV.js:1485
Shell command execution detected (child_process).
dist/shared-client-4ICy3U6d.js:93
Shell command execution detected (child_process).
Patterns worth reviewing
These patterns may indicate risky behavior. Check the VirusTotal and OpenClaw results above for context-aware analysis before installing.Verification
Tags
OpenClaw Codex
Official OpenClaw provider and harness plugin for OpenAI Codex app-server integration. It exposes the Codex-managed GPT model catalog and the Codex runtime surfaces used by OpenClaw agents.
Install from OpenClaw:
openclaw plugin add @openclaw/codex
Use this plugin when you want OpenClaw to run Codex-backed model turns, media understanding, and prompt overlays through the Codex app-server harness.