🧭 NEC Becomes Anthropic's First Japanese Global Partner — 30,000 Engineers to Adopt Claude Code
Anthropic and NEC Corporation announced a strategic collaboration on April 24 that positions NEC as the first Japan-based company to join Anthropic's global partner programme. The scope is substantial: all 30,000 NEC Group employees worldwide will gain access to Claude, and NEC has committed to building what it is calling Japan's largest AI-native engineering team — centred on Claude Code as the primary development tool.
What NEC is building with Anthropic
The partnership has three pillars. First, internal AI adoption: Claude will be rolled out to the 30,000-strong NEC workforce to accelerate joint development and deployment of AI solutions. NEC is establishing an internal Centre of Excellence staffed with highly skilled AI professionals, with technical support and training provided directly by Anthropic.
Second, product development: the two companies are beginning joint development of secure, industry-specific AI solutions for the Japanese market. The initial focus sectors are financial services, manufacturing, and local government — three verticals where NEC has existing deep relationships and enterprise install-base in Japan. Claude Cowork, Anthropic's desktop AI agent product, is named as the platform for these joint initiatives in the first phase.
Third, cybersecurity: NEC will integrate Claude into NEC BluStellar Scenario, the company's next-generation security service and its core value-creation model. This marks Claude's entry into NEC's flagship cybersecurity product line — a meaningful signal given how security-sensitive Japanese enterprise clients tend to be about AI tooling.
Why this matters for Claude Code adoption outside the US
NEC's stated ambition — building Japan's largest AI-native engineering team — is the clearest sign yet that Claude Code is becoming the reference tool for enterprise developer tooling outside the English-speaking market. NEC is a 119-year-old company with deep government, defence, and critical infrastructure relationships in Japan. Committing to Claude Code at scale is not a casual vendor choice.
What "AI-native engineering" means in practice
NEC's framing aligns with a pattern emerging at other large SI (systems integrator) firms: the goal is not simply to give engineers an AI assistant, but to reorganise engineering workflows around AI-first processes — where Claude Code handles implementation, code review, test generation, and documentation, and human engineers focus on architecture, requirements, and judgement calls. Anthropic's direct training support for NEC's Centre of Excellence suggests this is a multi-year capability-building commitment, not a license deal.
Context: the Japan AI market
Japan's enterprise AI market is large, conservative, and relationship-driven. NEC, alongside Fujitsu, Hitachi, and NTT Data, is one of the four dominant SI firms that together handle the majority of large Japanese enterprise and government IT programmes. Anthropic landing NEC as a first Japanese global partner gives Claude access to exactly the kind of channel that OpenAI and Google have been competing for. NEC's Nikkei Asia coverage of the tie-up described the deal as an attempt to "cultivate corporate AI demand" in Japan — suggesting NEC will actively resell Claude-based services to its client base, not just use Claude internally.
NEC
Japan
enterprise
Claude Code
AI-native engineering
partnership
Claude Cowork
cybersecurity
🧭 Claude Code v2.1.119: /config Persistence, GitLab/Bitbucket --from-pr, and Hooks duration_ms
Claude Code v2.1.119 landed at 23:24 on April 23 — effectively the morning of April 24 for most of the world's developer timezones — with a focused set of changes that improve settings management, multi-platform PR workflows, and hooks observability. Here is what changed.
/config settings now persist to settings.json
Previously, changes made via the /config command existed only for the current session. In v2.1.119, /config changes are written to ~/.claude/settings.json immediately, following the full project / local / policy override precedence chain. This means developer preferences set in /config (model, permission mode, thinking budget, etc.) survive restarts without requiring a manual edit to the settings file.
Practical impact of settings persistence
The change makes /config the recommended way to make persistent user-level settings changes, instead of having to find and edit ~/.claude/settings.json directly. One workflow this enables: use /config during a session to tune settings, and they will automatically be present in the next session without any file editing.
--from-pr now accepts GitLab, Bitbucket, and GitHub Enterprise URLs
The --from-pr flag — which initialises a Claude Code session with the full diff and context of an existing pull request — previously only accepted github.com URLs. v2.1.119 expands this to accept GitLab, Bitbucket, and GitHub Enterprise Server PR URLs. Teams not on github.com (a significant share of large enterprises, which typically run self-hosted GitLab or GHE) can now use --from-pr in the same way.
# Example: GitLab MR URL
claude --from-pr https://gitlab.yourcompany.com/myteam/myrepo/-/merge_requests/42
# Example: Bitbucket PR URL
claude --from-pr https://bitbucket.org/myteam/myrepo/pull-requests/17
# Example: GitHub Enterprise PR URL
claude --from-pr https://github.yourcompany.com/org/repo/pull/88
prUrlTemplate: custom code-review URL patterns
A new prUrlTemplate setting lets you define a custom URL template for code-review links that Claude Code surfaces after completing a PR-related task. Set it in settings.json or via /config:
# In ~/.claude/settings.json or project .claude/settings.json:
{
"prUrlTemplate": "https://git.yourcompany.com/{repo}/pull/{number}"
}
Useful for non-standard Git hosting platforms or internal review tools where the PR URL pattern differs from the platform defaults.
Hooks now report duration_ms
Every hook invocation now receives a duration_ms field in its event payload, recording how long the tool execution that triggered the hook took to complete. This is immediately useful for performance-sensitive hooks — for example, a PostToolUse hook that logs slow Bash commands, or a monitoring hook that alerts when a tool call exceeds an expected latency threshold.
# duration_ms in a PostToolUse hook payload
{
"tool_name": "Bash",
"tool_input": { "command": "npm test" },
"tool_result": { "output": "..." },
"duration_ms": 12430, # ← NEW in v2.1.119
"session_id": "sess_abc123",
"timestamp": "2026-04-24T06:15:22Z"
}
Other changes in v2.1.119
- CLAUDE_CODE_HIDE_CWD env var — suppress the current working directory display in the Claude Code header. Useful in screen-share or recording contexts where you don't want the local filesystem path visible.
- --print mode honours tools: / disallowedTools: frontmatter — agent CLAUDE.md files with a
tools: or disallowedTools: specification now correctly constrain --print non-interactive runs in addition to interactive sessions.
- PowerShell commands can be auto-approved in permission mode — previously, PowerShell tool calls on Windows required explicit approval even in permissive configurations. Fixed.
- Improved slash command suggestions — the
/ picker UI now shows better ranked completions based on recency and context.
Claude Code
v2.1.119
/config
--from-pr
GitLab
Bitbucket
hooks
duration_ms
settings.json