{"slug": "claude-code-v2-1-319-unblocks-unattended-agent-runs", "title": "Claude Code v2.1.319 unblocks unattended agent runs", "summary": "Anthropic released Claude Code v2.1.319, an update aimed at keeping long agent sessions running without human supervision. The release moves task-type MCP tool calls to the background so sessions no longer block on them, adds a denyInsteadOfPrompting permissions setting that denies unconfirmed tool calls instead of stalling, introduces a /claude-api latency-optimize command that proposes and eval-verifies faster API settings, and lets CLAUDE_CODE_AUTO_MODE_SERVER enable server-side auto mode checking even when experimental betas are disabled.", "body_md": "📅 November 6, 2026 • GitHub\n\n**Claude Code v2.1.319 targets unattended agent runs**\n\nAnthropic released Claude Code **v2.1.319**, and per the release notes the update concentrates on one operational problem - keeping long agent sessions moving when nobody is watching the terminal. The changes cover background execution for MCP tasks, a new permissions setting, a latency command for API code and a gateway setting for auto mode.\n\n**MCP tool calls stop blocking the session**\n\nMCP tool calls that run as tasks are now sent to the background immediately, and their result is reported on completion. A running Claude Code session no longer waits on them - the agent continues with the rest of its plan while a slow database query, a long-running build check or any other task-type MCP call finishes on its own. The result arrives as a completion report instead of a pause.\n\n**denyInsteadOfPrompting - approvals become policy**\n\nThe new permissions setting denyInsteadOfPrompting changes what happens at the moment a tool call would normally stop and wait for user confirmation. The call is denied instead, and the run continues - in the terminal, in subagents and in teammates. For scheduled or headless runs this replaces an indefinite stall with a defined outcome, and the session keeps producing until it reaches a natural stopping point.\n\n**Latency optimization as a command**\n\nA new /claude-api latency-optimize command finds places where your Claude API code causes waits and selects faster settings, verified against evals. Instead of manually profiling requests and guessing at parameter changes, developers get a command that proposes concrete adjustments and validates them against a measurable baseline.\n\n**Auto mode behind LLM gateways**\n\nTeams that route Claude Code through LLM gateways get a behavior change as well - CLAUDE_CODE_AUTO_MODE_SERVER now enables server-side checking even when CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS is set. Infrastructure that deliberately keeps experimental betas off retains a working auto mode path, which removes a configuration conflict for gateway operators.\n\n**IMPACT**\n\nThe affected audience is engineers who run Claude Code in CI pipelines, cron jobs, long autonomous coding sessions and multi-agent setups with subagents and teammates. Daily work changes in concrete ways. Unattended runs no longer freeze on an approval prompt - with denyInsteadOfPrompting the policy decides, and the run proceeds. Slow MCP tool calls no longer serialize a session - background execution keeps the plan moving while tasks complete. And latency tuning stops being a manual review exercise - the /claude-api latency-optimize command turns it into a repeatable step with eval-verified results. Gateway operators should check that auto mode behavior still matches their expectations after the update.\n\n**Sources:** github.com\n\n**Continue the conversation:** [https://t.me/AIAutomationArchitecture](https://t.me/AIAutomationArchitecture) · [Bluesky](https://bsky.app/profile/delimitter.bsky.social)", "url": "https://wpnews.pro/news/claude-code-v2-1-319-unblocks-unattended-agent-runs", "canonical_source": "https://dev.to/delimitter/claude-code-v21319-unblocks-unattended-agent-runs-476e", "published_at": "2026-11-06 20:08:13+00:00", "updated_at": "2026-11-06 20:23:51.122312+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "agent-protocols", "mlops"], "entities": ["Anthropic", "Claude Code", "MCP", "CLAUDE_CODE_AUTO_MODE_SERVER", "CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS", "GitHub"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/claude-code-v2-1-319-unblocks-unattended-agent-runs", "markdown": "https://wpnews.pro/news/claude-code-v2-1-319-unblocks-unattended-agent-runs.md", "text": "https://wpnews.pro/news/claude-code-v2-1-319-unblocks-unattended-agent-runs.txt", "jsonld": "https://wpnews.pro/news/claude-code-v2-1-319-unblocks-unattended-agent-runs.jsonld"}}