Casos de estudio, notas de versión y análisis técnicos
The v2.3–v2.4 release series closes the agent learning loop: a lesson promotion pipeline lets agents surface their own rule candidates, dynamic keyword-matched rule injection replaces the 58KB static payload, and slot-aware retry finally makes the taskboard reliable under load.
Leer →v2.2.4 hardens the runtime: Discord now retries sends and auto-recovers stale sessions, a worktree lock prevents Bash tool failures, dispatched tasks survive HTTP disconnects, and coordinator findings are no longer truncated.
Leer →How Tetora turns a single cron expression into a fully automated, human-gated workflow — and when to use each piece of the scheduling stack.
Leer →OpenClaw's April block-out had two root causes: ToS violations and open security exposure. Both were architecture choices, not accidents.
Leer →Anthropic's April 2026 enforcement clarified what's allowed with Claude subscriptions. Here's exactly how Tetora integrates with Claude — and why we won't get blocked.
Leer →Interactive model switcher, VibeVoice TTS, Human Gate improvements, Skill AllowedTools, and the Discord !model command.
Leer →DangerousOpsConfig blocks destructive commands before agents run them. Multi-tenant --client flag, worktree failure preservation, History CLI for failure analysis — v2.2 brings agent dispatch to production-grade.
Leer →Japanese CPA Kento Hatakeyama serves 60 companies solo with Claude Code. The system he spent months building maps perfectly to Tetora's out-of-the-box features.
Leer →256 files consolidated to a lean core. New Workflow Engine with DAG support and Template Marketplace.
Leer →