Skip to Content
Developer GuideMigrationOverview

Migration

Coming from another agent system? This section helps you transition to the Super Agents architecture.


Why Migrate?

If you’ve been using another Claude-based agent setup, you’ve already proven that LLM-assisted workflows work for you. The Super Agents architecture builds on that foundation with:

CapabilityTypical SetupSuper Agents
Context managementManual prompting or single config fileLayered context architecture with auto-refresh
MemoryIn-session only or basic DBPARA vault + structured knowledge base
CommandsFew or noneLarge slash-command library across domains
AgentsGeneric assistantSpecialized role-based personas
AutomationManual triggersLaunchAgents + scheduled pipelines
Thinking toolsNoneChallenge, discovery, planning, and learning tools
Multi-domainSingle-purposeUnified system across all life/work domains

Migration Paths

From OpenClaw / ClaudeClaw

If you’ve been using the Telegram-based OpenClaw/ClaudeClaw setup (Agent SDK subprocess architecture), see From OpenClaw.

Key changes: Telegram relay becomes optional, SQLite memory converts to vault files, single-purpose bot becomes multi-domain system.

From Scratch

No existing agent system? Start fresh with the From Scratch guide, which walks you through assessment and setup in a recommended order.

From Other Setups

If you’re migrating from a different system (custom GPT workflows, LangChain agents, AutoGPT, etc.), the general approach is:

  1. Audit what you have — list every workflow, command, and automation
  2. Map to Super Agents components — which become commands, agents, or skills?
  3. Migrate memory/context — convert any stored knowledge to vault markdown files
  4. Set up context loading — create your AGENT_GUIDE.md and /contxt bundles
  5. Migrate automations — convert scheduled tasks to LaunchAgents
  6. Verify — run Day 1 Verification to confirm everything works
Last updated on