[ca] chore: update CLI/MCP tool versions (Claude Code 2.1.126, Copilot 1.0.39, Codex 0.128.0, Playwright MCP 0.0.72, MCP Gateway v0.3 [Content truncated due to length]

[github-actions]

Summary

5 CLI/MCP tools have been updated in pkg/constants/version_constants.go, and make recompile has been run successfully (205/205 workflows recompiled).

Tool	Previous	New	Change
Claude Code	2.1.119	2.1.126	+7 patch
GitHub Copilot CLI	1.0.36	1.0.39	+3 patch
OpenAI Codex	0.125.0	0.128.0	+3 minor
Playwright MCP	0.0.70	0.0.72	+2 patch
MCP Gateway	v0.3.1	v0.3.3	+2 patch
GitHub MCP Server	v1.0.3	v1.0.3	no change
Playwright Browser	v1.59.1	v1.59.1	no change

---

Update Claude Code

• Version: 2.1.119 → 2.1.126
• Intermediate versions: 2.1.120, 2.1.121, 2.1.122, 2.1.123, 2.1.124, 2.1.126
• Note: No public GitHub repository. stable tag is pinned to 2.1.118; latest/next are at 2.1.126.

View Details

CLI Discovery

No public changelog available. Version info from NPM only.

Impact Assessment

• Risk: Low — patch-level bumps only
• Affects: All Claude engine agentic workflows

Package Links

• NPM Package: https://www.npmjs.com/package/`@anthropic-ai/claude-code`

---

Update GitHub Copilot CLI

• Version: 1.0.36 → 1.0.39
• Intermediate versions: 1.0.37, 1.0.38, 1.0.39
• Note: Repository is private. No public release notes available.

View Details

CLI Discovery

Package description: "GitHub Copilot CLI brings the power of Copilot coding agent directly to your terminal."

A prerelease 1.0.40-3 is available but not yet stable.

Impact Assessment

• Risk: Low — patch-level bumps
• Affects: All Copilot engine agentic workflows
• Verify: MCPs not blocked from loading; /models does not silently fail on PATs (as noted in code comment)

Package Links

• NPM Package: https://www.npmjs.com/package/`@github/copilot`

---

Update OpenAI Codex

• Version: 0.125.0 → 0.128.0
• Release Date: 2026-04-30
• Full Changelog: openai/codex@rust-v0.125.0...rust-v0.128.0

View Full Changelog

New Features

• Persisted /goal workflows — foundation, app-server API, model tools, runtime continuation, and TUI controls (create/pause/resume/clear): Add goal persistence foundation (1 / 5) openai/codex#18073, Add goal app-server API (2 / 5) openai/codex#18074, Add goal model tools (3 / 5) openai/codex#18075, Add goal core runtime (4 / 5) openai/codex#18076, Add goal TUI UX (5 / 5) openai/codex#18077, Use /goal resume for paused goals openai/codex#20082
• codex update command: Add codex update command openai/codex#19933
• Configurable TUI keymaps: feat(tui): add configurable keymap support openai/codex#18593
• Plan-mode nudges from composer drafts: feat(tui): suggest plan mode from composer drafts openai/codex#19901
• Action-required terminal titles: Show action required in terminal title openai/codex#18372
• /statusline and /title slash commands during active turns: Allow /statusline and /title slash commands during active turns openai/codex#19917
• Permission profiles — built-in defaults, sandbox CLI profile selection, cwd controls, active-profile metadata: permissions: add built-in default profiles openai/codex#19900, feat(cli): add explicit sandbox permission profiles openai/codex#20117, feat(cli): add sandbox profile config controls openai/codex#20118, permissions: expose active profile metadata openai/codex#20095
• Plugin improvements — marketplace installation, remote bundle caching, remote uninstall, plugin-bundled hooks, hook enablement state, external-agent config import: /plugins: add marketplace install flow openai/codex#18704, feat: Cache remote plugin bundles on install openai/codex#19914, Add remote plugin uninstall API openai/codex#19456, Discover hooks bundled with plugins openai/codex#19705, Add persisted hook enablement state openai/codex#19840, Support detect and import MCP, Subagents, hooks, commands from external openai/codex#19949
• External agent session import (including background import): External agent session support openai/codex#19895, Import external agent sessions in background openai/codex#20284, Consume ai-title from external sessions and add end marker openai/codex#20261
• MultiAgentV2 improvements — thread caps, wait-time controls, root/subagent hints, v2-specific depth: feat: surface multi-agent thread limit in spawn description openai/codex#19360, multi_agent_v2: move thread cap into feature config openai/codex#19792, Add MultiAgentV2 root and subagent context hints openai/codex#19805, Make MultiAgentV2 wait minimum configurable openai/codex#20052, Make multi-agent v2 ignore agents.max_depth openai/codex#20180

Bug Fixes

• Fix hang on turn/interrupt: Fix hang on turn/interrupt openai/codex#18392
• Restore persisted model provider on thread resume: Restore persisted model provider on thread resume openai/codex#19287
• Allow large remote app-server resume responses: Allow large remote app-server resume responses openai/codex#19920
• TUI reliability: terminal resize reflow, markdown list spacing, slash-command popup layout, keyboard cleanup, shell-mode escape: fix(tui): reflow scrollback on terminal resize openai/codex#18575, Preserve TUI markdown list spacing after code blocks openai/codex#19706, Keep slash command popup columns stable while scrolling openai/codex#19511, Reset TUI keyboard reporting on exit openai/codex#19625, fix(tui): let esc exit empty shell mode openai/codex#19986, Restore TUI working status after steer message is set openai/codex#19939
• Hardened managed network: deferred denials, proxy bypass defaults, resolved target checks, IPv6 host matching, git -C approval: fix: handle deferred network proxy denials openai/codex#19184, fix(network-proxy): tighten network proxy bypass defaults openai/codex#20002, fix(network-proxy): recheck network proxy connect targets openai/codex#19999, fix(network-proxy): normalize network proxy host matching openai/codex#19995, fix: don't auto approve git -C ... openai/codex#20085
• Windows sandbox and PTY edge cases: Fix Windows pseudoconsole attribute handling for sandboxed PTY sessions openai/codex#20042, Improve Windows process management edge cases openai/codex#19211, expand the set of core shell env vars for Windows. openai/codex#20089, check PID of named pipe consumer openai/codex#19283
• Bedrock model apply_patch, GPT-5.4 reasoning levels: Fix: use function apply_patch tool for Bedrock model openai/codex#19416, fix: Bedrock GPT-5.4 reasoning levels openai/codex#19461, feat: update Bedrock Mantle endpoint and GPT-5.4 model ID openai/codex#20109
• MCP/plugin edge cases: stdio server cleanup, plugin MCP approval persistence, custom MCP metadata isolation: Terminate stdio MCP servers on shutdown to avoid process leaks openai/codex#19753, [mcp] Fix plugin MCP approval policy. openai/codex#19537, disallow fileparams metadata for custom mcps openai/codex#19836, Strip connector provenance metadata from custom MCP tools openai/codex#19875

Deprecations

• --full-auto flag deprecated; users should use explicit permission profiles: chore(cli) deprecate --full-auto openai/codex#20133
• Stop publishing GNU Linux release artifacts: ci: stop publishing GNU Linux release artifacts openai/codex#19445

Impact Assessment

• Risk: Medium — significant feature additions (permission profiles, plugin marketplace, codex update)
• Affects: Codex engine agentic workflows
• Notable: Permission profile system is now more explicit; --full-auto deprecated

Package Links

• NPM Package: https://www.npmjs.com/package/`@openai/codex`
• Repository: https://github.com/openai/codex
• Release Notes: https://github.com/openai/codex/releases/tag/rust-v0.128.0

---

Update Playwright MCP

• Version: 0.0.70 → 0.0.72
• Note: Follows Playwright browser versioning patterns.

View Details

CLI Discovery

No public changelog for individual patch releases. NPM description: "Playwright Tools for MCP"

Impact Assessment

• Risk: Low — minor patch bumps
• Affects: Playwright browser automation workflows

Package Links

• NPM Package: https://www.npmjs.com/package/`@playwright/mcp`
• Repository: https://github.com/microsoft/playwright

---

Update MCP Gateway

• Version: v0.3.1 → v0.3.3 (via v0.3.2)
• Docker Image: ghcr.io/github/gh-aw-mcpg:v0.3.3

View v0.3.2 Changelog

Bug Fixes

• ARC/DinD runner support — DOCKER_HOST propagation, TCP socket detection, Kubernetes cgroup container ID detection: fix: MCP gateway ARC/DinD support — DOCKER_HOST propagation, TCP socket, K8s cgroup ID detection gh-aw-mcpg#4814
• Data race fix in argument validation canary test using atomic.Bool: go-fan follow-up: argument validation canary, MaxRetries -1 sentinel docs, isSessionNotFoundError TODO gh-aw-mcpg#4811
• Middleware logging — correct chmod label, init-time jq compile error visibility: Remove duplicate logMiddleware.Printf calls in jqschema middleware gh-aw-mcpg#4809

Performance

• Native Go walk_schema — replaced recursive jq-based schema walker with native Go, improving throughput and eliminating CGo overhead: feat(middleware): replace recursive jq walk_schema with native Go implementation gh-aw-mcpg#4750
• Eliminate os.Stat syscall and json.Marshal in large-payload middleware hot path: [Repo Assist] perf(middleware): eliminate os.Stat syscall and json.Marshal in large-payload path gh-aw-mcpg#4788

Refactors

• Split large monolithic files into focused modules: refactor: split large monolithic files identified by semantic clustering analysis gh-aw-mcpg#4813
• Deduplicate guard init pipeline, DIFC component fields, HTTP error helpers: refactor: deduplicate guard init pipeline, DIFC component fields, and HTTP error helpers gh-aw-mcpg#4740
• Extract copySchemaWithKey helper: refactor: extract copySchemaWithKey helper to deduplicate inline map-copy pattern in NormalizeInputSchema gh-aw-mcpg#4810

18 merged PRs since v0.3.1 — see https://github.com/github/gh-aw-mcpg/releases/tag/v0.3.2

View v0.3.3 Changelog

Bug Fixes

• Routed mode session timeout — Replace hardcoded 30-min timeout with configurable MCP_GATEWAY_SESSION_TIMEOUT (6h default), fixing "session not found" errors in long-running agentic workflows: Fix "session not found" errors in routed mode for long-running agentic workflows gh-aw-mcpg#4849
• Auth error handling — Gracefully handle authentication errors in the MCP Gateway Log Analyzer: fix: handle auth errors gracefully in MCP Gateway Log Analyzer gh-aw-mcpg#4878

Tracing

• HTTP response status on spans — Record HTTP status codes on OTel spans, sanitize error recording, bump semconv to v1.27.0: feat(tracing): HTTP response status on spans, sanitize error recording, bump semconv to v1.27.0 gh-aw-mcpg#4879

Refactors

• Rust guard constants — add scope_names constants: [Repo Assist] refactor(rust-guard): add scope_names constants and replace bare string literals gh-aw-mcpg#4857
• Logger cleanup — replace manual logger close calls with CloseAllLoggers(): Replace manual logger close calls in postRun with CloseAllLoggers() gh-aw-mcpg#4881

9 merged PRs since v0.3.2 — see https://github.com/github/gh-aw-mcpg/releases/tag/v0.3.3

Impact Assessment

• Risk: Low-Medium — important session timeout fix benefits long-running workflows
• Affects: All agentic workflows using MCP Gateway (routed mode)
• Notable: Session timeout is now configurable via MCP_GATEWAY_SESSION_TIMEOUT (default 6h instead of 30min); critical fix for long-running workflows

Package Links

• Repository: https://github.com/github/gh-aw-mcpg
• Release v0.3.2: https://github.com/github/gh-aw-mcpg/releases/tag/v0.3.2
• Release v0.3.3: https://github.com/github/gh-aw-mcpg/releases/tag/v0.3.3

---

Recommendations

1. MCP Gateway (v0.3.1→v0.3.3): High priority — session timeout fix directly impacts long-running agentic workflow reliability
2. Codex (0.125.0→0.128.0): Medium priority — large feature release; validate permission profile changes and --full-auto deprecation
3. Claude Code (2.1.119→2.1.126): Low priority — patch updates with no public changelog
4. Copilot CLI (1.0.36→1.0.39): Low priority — verify MCP loading and PAT auth still work
5. Playwright MCP (0.0.70→0.0.72): Low priority — minor patch bumps

References:

• §25208776945

Generated by CLI Version Checker · ● 320.3K · ◷

• expires on May 3, 2026, 9:03 AM UTC