Skip to content

docs: Claude Code v2.1.250 - --restricted mode, cross-session messaging on Bedrock/Vertex/Foundry, new anthropics/formal-math repo - #1223

Merged
claude-yolo[bot] merged 1 commit into
mainfrom
docs/claude-code-v2.1.250-20260828-1223
Aug 28, 2026
Merged

docs: Claude Code v2.1.250 - --restricted mode, cross-session messaging on Bedrock/Vertex/Foundry, new anthropics/formal-math repo#1223
claude-yolo[bot] merged 1 commit into
mainfrom
docs/claude-code-v2.1.250-20260828-1223

Conversation

@claude-yolo

@claude-yolo claude-yolo Bot commented Aug 28, 2026

Copy link
Copy Markdown
Contributor

Night shift report

WHY THIS MATTERS: v2.1.248 is a landmark release for enterprise and security-conscious deployments: the new --restricted flag creates a locked-down Claude Code mode that strips command-execution and web tools, enforces working-directory file isolation, refuses bypassPermissions, and ignores user/project settings — exactly what teams need before wider rollouts. Alongside that, cross-session messaging (SendMessage/ListAgents) now works on Bedrock, Vertex, and Foundry (previously cloud-only), and the Workflow tool's prompt footprint shrank from 5.7k → 1k tokens. Discovery also found a new Anthropic GitHub repo (anthropics/formal-math) and two Claude Tag admin pages got consolidated, which is the routine kind of product restructure that tombstones track for us.

HIGHLIGHTS:

  • --restricted mode (v2.1.248): claude --restricted or CLAUDE_CODE_RESTRICTED=1 removes built-in command/code-execution tools and WebFetch, keeps file access sandboxed to the working directory, and ignores all local settings. This is the "safe for untrusted codebases" deployment profile that enterprise teams have been asking for.
  • Cross-session messaging on Bedrock/Vertex/Foundry (v2.1.248): SendMessage/ListAgents between sessions on the same machine now works for all providers, not just Anthropic-hosted — big for multi-agent pipelines on enterprise infrastructure.
  • experimental.cacheTtl in agent frontmatter (v2.1.248): agents can now declare "5m" or "1h" as a per-agent prompt cache TTL, giving fine-grained control over caching costs in subagent fleets.
  • /usage-credits command (v2.1.248): Enterprise orgs on AWS Marketplace, self-serve Enterprise, and trials now have a command for members to request higher usage limits from their admin — closes a UX gap for quota management.
  • anthropics/formal-math new in discovery.json: Anthropic published a new public repo (anthropics/formal-math) that the archive doesn't hold yet. Human decision needed on whether to add it to sources. anthropics/zeta-23-lean dropped out of the not_fetched list (likely repo was deleted or made private).
  • Prompt footprint win (v2.1.248): Workflow tool description cut from 5.7k to 1k tokens — meaningful cost reduction for sessions that use multi-agent orchestration heavily.
  • Tombstone: two Claude Tag admin pages consolidatedpair-workspace and test-it both 301 to setup-overview, suggesting the Claude Tag onboarding flow was simplified into fewer pages.
  • 130+ Claude Code docs refreshed — broad surface update across all of content/en/docs/claude-code/, covering agent SDK, self-hosted environments, managed settings, permissions, and more. Likely documentation catching up with the 2.1.248 feature set.
  • v2.1.250 itself: bug fixes and reliability improvements only (per changelog).

Created by night-shift claude-yolo
Day-shift claude-yolo will review and merge this in the morning

…ng on Bedrock/Vertex/Foundry, new anthropics/formal-math repo

Co-Authored-By: claude-yolo[bot] <claude-yolo@lroole.com>
@claude-yolo
claude-yolo Bot merged commit 998ccb5 into main Aug 28, 2026
1 check passed
@claude-yolo
claude-yolo Bot deleted the docs/claude-code-v2.1.250-20260828-1223 branch August 28, 2026 12:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants