Summary
Two agentic CLI tools have new stable releases. Constants in pkg/constants/version_constants.go have been updated and make recompile regenerated 217 workflow lock files with the new versions.
| Tool |
Current |
New |
Status |
| Claude Code |
2.1.165 |
2.1.167 |
⬆️ Updated |
| Copilot CLI |
1.0.59 |
1.0.60 |
⬆️ Updated |
| Codex |
0.137.0 |
0.137.0 |
✅ Current |
| GitHub MCP Server |
v1.1.2 |
v1.1.2 |
✅ Current |
| MCP Gateway |
v0.3.23 |
v0.3.23 |
✅ Current |
| Playwright MCP |
0.0.75 |
0.0.75 |
✅ Current |
| Playwright CLI |
0.1.13 |
0.1.13 |
✅ Current |
| Playwright Browser |
v1.60.0 |
v1.60.0 |
✅ Current |
Update Claude Code
- Previous: 2.1.165 → New: 2.1.167 (intermediate: 2.1.166; 2.1.164 was unpublished/skipped)
- Timeline: 2.1.166 (2026-06-05), 2.1.167 (2026-06-06)
- Risk: Low — patch releases only
Claude Code has no public repository; analysis relies on NPM metadata and CLI help output. The claude --help output is identical between 2.1.162 and 2.1.167, confirming these are pure patch releases with no user-facing CLI surface changes.
Update Copilot CLI
- Previous: 1.0.59 → New: 1.0.60
- Timeline: 1.0.60 (2026-06-05)
- Risk: Low–Medium — one Windows-only breaking change (see below)
Key Features
- Reasoning effort levels: max reasoning effort for Anthropic models, all levels on every plan
- Git worktrees from PRs: create a worktree for a PR directly from the pull requests screen
- New
-r shorthand for --resume
- New
--enable-memory flag (memory in prompt mode, disabled by default)
--model auto lets Copilot pick the model automatically- New
billing help topic (AI Credit Usage)
- Vim-style navigation (g, G, Ctrl+D, Ctrl+U) in
/diff; Ctrl+S now stashes/pops the prompt
- New config
builtInAgents.rubberDuckAutoInvoke (default false); new models gemini-3.1-pro-preview, gemini-3.5-flash
Security Fixes
web_fetch now blocks loopback, private, and cloud-metadata addresses and no longer silently follows redirects- Prompts for approval before commands that set dynamic-loader/git-config env vars (e.g.
LD_PRELOAD)
- Rewind no longer deletes ignored files when rolling back
Breaking Change
- Windows path discovery: executables are no longer discovered in the working directory when invoked by bare name (e.g.
git). Users must add the working directory to PATH to restore discovery. (Not expected to affect Linux CI runners.)
CLI help diff (1.0.59 → 1.0.60)
Main --help:
- added
--enable-memory
--model <model> now documents auto- added
-r shorthand for --resume
- added
billing help topic
config subcommand:
- new models:
gemini-3.1-pro-preview, gemini-3.5-flash
- new setting:
builtInAgents.rubberDuckAutoInvoke (default false)
environment subcommand: no changes.
Upgrade verification checklist (per constants.go note)
No-change tools
Codex 0.137.0, GitHub MCP Server v1.1.2, MCP Gateway v0.3.23, Playwright MCP 0.0.75, Playwright CLI 0.1.13, and Playwright Browser v1.60.0 are all already current. (MCP Gateway was bumped v0.3.22→v0.3.23 in a prior change before this run.)
Recompile note
make recompile regenerated 217 lock files successfully with the new versions. 5 workflows fail strict-mode compilation due to a pre-existing deprecated max-effective-tokens field (unrelated to this version bump): schema-consistency-checker, workflow-health-manager, daily-safe-output-optimizer, agent-performance-analyzer, pr-triage-agent. Their lock files retain the old versions until that deprecated field is removed from the source .md files.
Recommendations
- Merge the constants + lock file updates (Low risk).
- Separately, remove the deprecated
max-effective-tokens field from the 5 workflows above so they can recompile onto the new versions.
Package Links
References: §27055099961
Generated by 🔢 CLI Version Checker · 139.8 AIC · ⌖ 25.1 AIC · ◷
Summary
Two agentic CLI tools have new stable releases. Constants in
pkg/constants/version_constants.gohave been updated andmake recompileregenerated 217 workflow lock files with the new versions.Update Claude Code
Claude Code has no public repository; analysis relies on NPM metadata and CLI help output. The
claude --helpoutput is identical between 2.1.162 and 2.1.167, confirming these are pure patch releases with no user-facing CLI surface changes.Update Copilot CLI
Key Features
-rshorthand for--resume--enable-memoryflag (memory in prompt mode, disabled by default)--model autolets Copilot pick the model automaticallybillinghelp topic (AI Credit Usage)/diff; Ctrl+S now stashes/pops the promptbuiltInAgents.rubberDuckAutoInvoke(defaultfalse); new modelsgemini-3.1-pro-preview,gemini-3.5-flashSecurity Fixes
web_fetchnow blocks loopback, private, and cloud-metadata addresses and no longer silently follows redirectsLD_PRELOAD)Breaking Change
git). Users must add the working directory toPATHto restore discovery. (Not expected to affect Linux CI runners.)CLI help diff (1.0.59 → 1.0.60)
Main
--help:--enable-memory--model <model>now documentsauto-rshorthand for--resumebillinghelp topicconfigsubcommand:gemini-3.1-pro-preview,gemini-3.5-flashbuiltInAgents.rubberDuckAutoInvoke(defaultfalse)environmentsubcommand: no changes.Upgrade verification checklist (per constants.go note)
/modelsdoes not silently fail on PATsNo-change tools
Codex 0.137.0, GitHub MCP Server v1.1.2, MCP Gateway v0.3.23, Playwright MCP 0.0.75, Playwright CLI 0.1.13, and Playwright Browser v1.60.0 are all already current. (MCP Gateway was bumped v0.3.22→v0.3.23 in a prior change before this run.)
Recompile note
make recompileregenerated 217 lock files successfully with the new versions. 5 workflows fail strict-mode compilation due to a pre-existing deprecatedmax-effective-tokensfield (unrelated to this version bump):schema-consistency-checker,workflow-health-manager,daily-safe-output-optimizer,agent-performance-analyzer,pr-triage-agent. Their lock files retain the old versions until that deprecated field is removed from the source.mdfiles.Recommendations
max-effective-tokensfield from the 5 workflows above so they can recompile onto the new versions.Package Links
References: §27055099961