Changelog

Track shipped product updates, release notes, and major experience changes in Tulsk.

Latest Release

Version 2.1.0 — Multi-Agent Clusters & EMA as PM

Multi-Agent OpenClaw Clusters

• Run up to 6 AI agents on a single managed VPS for $39/mo
• Each agent gets its own isolated workspace with IDENTITY.md, SOUL.md, USER.md, and TOOLS.md
• Agents on the same cluster can communicate directly via OpenClaw’s native agent-to-agent protocol
• Native subagent support — agents can spawn temporary workers to parallelize complex tasks
• Cluster management UI: create, reprovision, stop, start, delete clusters from the agents page
• Per-agent persona editor with tabs for IDENTITY, SOUL, USER, and TOOLS
• Real-time heartbeat monitoring with line chart on cluster settings page

EMA as Project Manager

• EMA now acts as your AI Project Manager overseeing all agents
• System prompt automatically enriched with recent agent activity (last 10 completed runs)
• New get_agent_activity_summary tool — EMA can query detailed stats grouped by agent and project
• Agent results push back to EMA chat — trigger an agent from chat, result appears in the conversation automatically
• EMA proactively shares updates, connects dots between agent outputs, and recommends delegation

Scheduled Reports

• Tell EMA “Report every Monday about project progress” and it creates an automated schedule
• New EMA settings page at /settings/ema for managing scheduled reports
• Cron-based execution with configurable timezone
• Reports generated using LLM with full agent activity context, delivered to your chat

Agent-to-Agent Delegation

• Agents can delegate work to each other with full audit trails
• Delegation depth tracking (max 3 levels) with self-delegation prevention
• New get_delegation_tree tool — EMA can trace the entire delegation chain
• parentRunId and delegationDepth tracked on every AgentRun

Streaming Optimization

• Bridge now publishes streaming chunks directly to Redis (Upstash REST API)
• Skips the Tulsk API hop — reduces streaming latency from ~300ms to ~50ms per chunk

Other Improvements

• EMA avatar replaces robot icon in chat
• OpenClaw page updated with multi-agent features and 7-language translations
• Cluster settings page with server status, health, LLM model, billing, action buttons
• Agent list groups agents by cluster with compact card layout
• Agent deactivate/activate option in dropdown menu
• Truncated response detection — shows note when LLM output token limit is reached

Version 2.0.0

• Introduced autonomous agent workflows for GTM execution
• Expanded multilingual marketing support
• Improved collaboration across projects and tasks

Version 1.0.0

• Initial release of Tulsk