[ca] CLI Updates Available: Claude Code 2.1.169, Codex 0.138.0, GitHub MCP Server v1.2.0 #38060

Summary

Three monitored CLI/MCP tools have newer releases available (all published 2026-06-08):

No changes for: Copilot CLI (1.0.60), Copilot SDK (1.0.0), MCP Gateway (v0.3.25), Playwright MCP (0.0.75), Playwright CLI (0.1.13), Playwright Browser (v1.60.0).

All three are constant-only edits in pkg/constants/version_constants.go. CLI --help output is byte-identical between old and new versions for Claude Code and Codex (no command/flag changes).

Update Claude Code

• Previous: 2.1.168 → New: 2.1.169 (published 2026-06-08)
• No public repository; analysis based on NPM metadata and CLI help output.
• CLI Discovery: claude --help output is identical to 2.1.168 — no new commands or flags. Patch release.
• Risk: Low (single patch increment, no surface change)

Update OpenAI Codex

• Previous: 0.137.0 (2026-06-04) → New: 0.138.0 (2026-06-08)
• CLI Discovery: codex --help output identical to 0.137.0 — no top-level flag changes.

Key Features

• /app command hands off the CLI thread into Codex Desktop on macOS and native Windows; Windows workspace launches can open directly into Desktop (feat(tui): add /app desktop handoff openai/codex#25638, Open Windows app workspaces via deep link openai/codex#26500)
• Local image attachments and standalone image generations expose saved file paths to the model for more reliable follow-up edits (Expose local image paths to models openai/codex#25944, Add saved image path hint to standalone image generation openai/codex#25947)
• More flexible reasoning-effort selection: TUI fallback shortcuts for terminals without Alt bindings; model-defined effort levels flow through in model-advertised order (fix(tui): add reasoning effort fallback shortcuts openai/codex#25623, [codex] Support model-defined reasoning efforts openai/codex#26444, [codex] Use model-advertised reasoning effort order openai/codex#26446)
• App-server can read account token usage; auth supports v2 personal access tokens in CLI and app-server flows (feat(app-server): expose account token usage [1 of 2] openai/codex#25344, [codex-rs] support v2 personal access tokens openai/codex#25731)
• Richer plugin automation: --json for add/remove and marketplace commands; plugin list JSON includes marketplace source; plugin detail exposes default prompts, remote MCP servers, and unavailable app templates (Add JSON output for plugin subcommands openai/codex#26631, Expose configured marketplace source in plugin list JSON openai/codex#26417, Preserve remote plugin default prompts openai/codex#25887, fix(app-server): expose remote MCP servers in plugin read openai/codex#26453, [codex] Expose unavailable app templates in plugin detail openai/codex#26317)

Impact Assessment

• Risk: Low. No CLI flag changes; gh-aw invokes Codex non-interactively. Changes are largely TUI/desktop/plugin features irrelevant to headless workflow runs.

Update GitHub MCP Server

• Previous: v1.1.2 → New: v1.2.0 (published 2026-06-08)
• Docker image: ghcr.io/github/github-mcp-server:v1.2.0

Key Features

• New create_project and create_iteration_field methods in projects_write (feat: add create_project and create_iteration_field methods to projects_write github-mcp-server#2232)
• New include_patch parameter for the get_commit tool (feat: add include_patch parameter to get_commit tool github-mcp-server#1924)
• New confidence parameter on issue mutation tools (Add confidence parameter to issue mutation MCP tools github-mcp-server#2605)
• MCP Apps aligned with the 2026-01-26 stable spec (MCP Apps: align with the 2026-01-26 stable spec github-mcp-server#2573); created issue/PR links open via host open-link capability (MCP Apps: Open created issue/PR link via host open-link capability github-mcp-server#2593); MCP App form skipped when issue/PR write carries non-form params (Skip MCP App form when issue/PR write carries non-form params github-mcp-server#2589)

Impact Assessment

• Risk: Low. The gh-aw GitHub MCP server is read-only; new tools are additive (project/iteration writes, include_patch, confidence) and do not remove or change existing read tool signatures. The get_review_comments thread-id restoration is a positive fix for review-reading workflows.

Migration Guide

Apply these constant edits in pkg/constants/version_constants.go:

const DefaultClaudeCodeVersion Version = "2.1.169"        // was 2.1.168
const DefaultCodexVersion Version = "0.138.0"             // was 0.137.0
const DefaultGitHubMCPServerVersion Version = "v1.2.0"    // was v1.1.2

Then regenerate lock files (GitHub MCP is a Docker-pinned image, so a double recompile is recommended to refresh container SHA pins):

make build && make recompile && make recompile
git status   # verify *.lock.yml changes; do NOT commit pkg/workflow/js/*.js

Recommendations

• All three updates are Low risk and safe to batch in a single PR.
• Priority: routine. No security advisories surfaced in the release notes.
• Testing: a standard make recompile + existing CI is sufficient; no flag/behavior migration required.

References:

• https://www.npmjs.com/package/`@anthropic-ai/claude-code`
• https://github.com/openai/codex/releases/tag/rust-v0.138.0
• https://github.com/github/github-mcp-server/releases/tag/v1.2.0

Generated by 🔢 CLI Version Checker · 123.1 AIC · ⌖ 16.7 AIC · ⊞ 11.3K · ◷

• expires on Jun 10, 2026, 10:43 PM UTC-08:00