• Automated Email Intake — Incoming support emails are picked up automatically. Customers receive an immediate acknowledgment with a unique case ID — no manual triage needed.
• Parallel Agent Analysis — Four AI agents work simultaneously: one classifies the request, one retrieves contract details, one looks up applicable rules, and a fourth assembles the full picture. All within seconds.
• Ready-to-Act Case Files — Every case arrives fully prepared — classification, contract context, relevant policies, and a summary with recommended next steps. Support staff can start resolving immediately.
• Searchable Case History — Every processed case is stored automatically, building a searchable history over time. Spot patterns, track recurring issues, and review past resolutions.
• Support Agent Chat — Your team gets an AI-powered chat with full knowledge of all cases. Ask about a specific case, query trends, or get recommendations for handling a follow-up.

• Mail Reaktor Registration — Uses registerMailReaktorDefinition to create an email-triggered workflow. Incoming emails are automatically parsed into subject, body, sender, and message ID.
• Agent Encapsulation with defineAktor — Each AI agent is defined as a reusable defineAktor subgraph that encapsulates its own prompt template, message construction, and AI call. This makes agents independently testable and configurable.
• Automatic Parallel Execution — The Reaktor framework analyzes data dependencies in the Aktor graph. Agents 1, 2, and 3 have no shared inputs, so they execute in parallel. Agent 4 depends on Agents 2 and 3, so it waits automatically — no manual orchestration code needed.
• Template-Based Prompting — All agent prompts use aktorCompletePrompt with {{variable}} placeholders, filled at runtime with actual email content and agent outputs. Prompts are exposed as aktorSetting values, so non-developers can customize agent behavior through the UI.
• Cross-Reaktor State via Database — The Mail Extraction Reaktor writes cases; the Support Chat Reaktor reads them. Both share the same AgentDB instance configured via a deployment setting — no direct coupling between Reaktors.
• Dependency Management with aktorPassthrough — Ensures the database write completes before the reply email is sent, even though the write isn't in the direct return path of the Aktor graph.

The Multi-Agent Support system consists of two Reaktors that share an AgentDB. The Mail Extraction Reaktor is the core of the system: when an email arrives, it constructs an Aktor graph where four AI agents form the analysis layer. The framework's lazy evaluation engine detects that three of the four agents are independent and runs them in parallel. The Summary Agent's inputs depend on the Contract and Knowledge agents, so it waits for their results automatically. After all agents complete, the case is persisted to the database and a templated auto-reply is sent to the customer. The Support Chat Reaktor is simpler — it loads all cases from the shared database, injects them into a system prompt via aktorCompletePrompt, and provides a chat interface powered by aktorAICall where human agents can query case data conversationally.