aistack

Gate risky agent actions behind a human-in-the-loop consensus checkpoint using aistack's consensus tools — propose an action, pause for approval, and only proceed once approved. Use before high-impact or irreversible operations (deploys, schema migrations, mass edits, external side effects), or when the user wants approval gates / sign-off before agents act.

Use aistack's shared agent memory — store, search, and recall facts, decisions, and artifacts across agents and sessions via the aistack MCP server's memory tools. Use when work needs to persist context between agents or across sessions, when the user asks to "remember", "recall", or "what did we decide about X", or to avoid re-deriving information already established earlier.

Orchestrate a multi-agent run with aistack — pick the right specialized agent (coder, architect, reviewer, researcher, tester, security-auditor, etc.), spawn it via the aistack MCP server, and coordinate handoffs. Use when a task is large enough to benefit from decomposition across specialized agents, or when the user asks to "orchestrate", "spawn an agent", or "run a multi-agent workflow".