Maintainer of this skill?
Automated security scan — not a certification.
by SeemSeam
Visible multi-agent CLI workspace for mixing Codex, Claude, Gemini, Kimi, Qwen, Cursor, Copilot, Pi, OpenCode, and other AI coding agents
# Add to your Claude Code skills
git clone https://github.com/SeemSeam/claude_codex_bridgeGuides for using ai agents skills like claude_codex_bridge.
Last scanned: 5/10/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-05-10T06:32:58.802Z",
"semgrepRan": false,
"npmAuditRan": true,
"pipAuditRan": true
}See how claude_codex_bridge compares with popular alternatives.
claude_codex_bridge is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by SeemSeam. Visible multi-agent CLI workspace for mixing Codex, Claude, Gemini, Kimi, Qwen, Cursor, Copilot, Pi, OpenCode, and other AI coding agents. It has 3,609 GitHub stars.
Yes. claude_codex_bridge passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/SeemSeam/claude_codex_bridge" and add it to your Claude Code skills directory (see the Installation section above).
claude_codex_bridge is primarily written in Python. It is open-source under SeemSeam on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh claude_codex_bridge against similar tools.
No comments yet. Be the first to share your thoughts!
From $49 · 30 days · terms & refunds
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
Automated security scan — not a certification.
See comparison
A lightweight multi-agent TUI with a stable cross-provider collaboration layer Coordinate Codex, Claude, Gemini, and other CLI agents in visible, controllable workflows you can take over
中文 | English | 日本語 | Français | Deutsch | العربية | Español | Português | 한국어 | Русский
Quick Start · Message Queues · Mobile App · Rich Mode · Configure Agents · User Guide · Developer Guide
A -> B -> C, A,B -> C, and A -> B,C.OpenCode 2.2.x gains opt-in native task submission and idle completion. Legacy OpenCode reads use private snapshots to avoid touching Provider WAL indexes; OMP tasks use native ingress. See bilingual release notes and the OpenCode configuration contract.
OpenCode 2.2.x 新增显式启用的原生任务提交与完成判定；旧版 OpenCode 使用私有快照读取，OMP 使用原生任务入口。启用方式与限制见版本说明。
Chat and content readers share a selectable gpt_markdown renderer with syntax-highlighted code, copy controls, task checkboxes and LaTeX. Existing artifact routing and external-link confirmation are preserved. Bilingual notes.
OMP Mobile now renders native OMP Collab select/editor questions as interactive cards and returns answers through the fenced native ui-response channel over HTTP or encrypted Relay. The workspace also shows bounded native runtime status such as the current step, elapsed time, queue depth, model/thinking/context usage, task progress and Todo counts. Upgrade the server and app together; idle existing OMP processes must be restarted normally to load the bridge, and busy agents must not be interrupted.
OMP Mobile now paginates stable long history through bounded native parent-chain pages. History-query errors preserve a healthy WebSocket and already-loaded content. Oversized replies keep a structured latest-content bubble capped at 24 KiB, 8 entries and 24 blocks; complete originals remain available as a secondary reader. Upgrade both the server and app, then restart existing OMP processes normally when idle; do not interrupt busy agents or discard history.
OMP replies use bounded, owner-bound native reply_deliver / reply_status delivery. A ccb-reply is received as an aside at a safe step boundary; only a persisted native custom_message counts as a receipt and releases the mailbox. Receipt does not mean subsequent business processing is finished. Drafts and active ask anchors remain intact, no concurrent turn is started, and ambiguous sends are queried rather than replayed. Notes.
When CCB environment variables are filtered, OMP Python ask now recovers the sender identity from the exact configured managed PI_CODING_AGENT_DIR. Existing identity precedence is preserved, and historical jobs are not replayed. The WeChat group 2 QR code has also been refreshed and is valid until October 29. Notes.
Linux OMP Mobile now preserves earlier reply text while appending same-session updates, with tool reasoning folded in place. Runtime sends use a persisted FIFO and run serially; queued sends remain visible but cannot steer an active turn, and disconnects with an unknown outcome are held instead of blindly replayed. This release also includes the 8.7.14 Relay and large-history fixes. The legacy macOS route is unchanged. Notes.
Linux OMP agents now connect Mobile to the existing native TUI through a private loopback Collab relay, with live output, reasoning, tool calls, and results grouped into user-anchored reply bubbles. Mobile can select the current native model and supported thinking level in the same host session. The new route is Linux-only: macOS keeps the original OMP terminal-input and native-history route, and other providers are unchanged. Notes.
Recover proven stale delivery leases without duplicating pending jobs, correct silent/callback completion status, and recognize newer Claude/Codex input layouts. Notes.
Reuse verified plugin snapshots, share resolved dependencies, recover interrupted builds and honor startup timeout overrides. Verified with real Pi plugin loading and two-agent cold/warm startup. Notes.
Fix accepted jobs waiting behind native Codex fullscreen status/shortcut rows and ordinary replies mentioning esc to interrupt, while preserving human drafts, busy/menu guards and FIFO. Notes.
Recognize native Claude pasted envelopes and empty Codex composers with newer status bars; preserve draft protection. Notes.
Opening /model or another control-only selector no longer supersedes an
active OMP/Pi ask. CCB now waits for a real provider turn boundary before
classifying different user input as replacement work. See the
verification and upgrade guidance.
Claude and Codex preserve usable local history across account/key/gateway changes. Pi follows manual session switches; OMP accepts its newer history header format. See verified scope and upgrade guidance.
Codex composer detection no longer matches model names. Claude suggestion, accepted-draft and ordered return-delivery regressions accompany real continuous communication verification. Experimental Claude native plugins remain disabled. Install/update onboarding no longer recommends the six V3 preview roles; explicit installation and updates for installed roles remain available. See upgrade guidance and layout limitations.
Fix Codex recovery commands with duplicate resume arguments, abandoned OMP editor sockets, and padded Codex empty-input detection. Queue details now show input-guard waiting reasons. See upgrade guidance and remaining limitations.