Architecture
How Agent as a Backend processes every request
Request Flow
Frontend
User sends a natural language intent
Auth & Security
JWT validation, PALADIN injection defense
Intent Router
Kernel classifies intent via small model (<10ms)
Capability Registry
Dispatches to the matching registered Capability
Context Bus
Progressive memory hydration across 5 tiers
Worker Fleet
Typed workers execute with LiteLLM backend
AG-UI Event Stream
13 event types streamed via SSE to frontend
Mem0 Consolidation
Async fact extraction: ADD / UPDATE / DELETE / NOOP
Memory Tiers
Protocol Triangle
AG-UI
User-facing
SSE event streaming to frontends
MCP
Tool-facing
Aggregate external tools or expose agent as tool
A2A
Agent-facing
Agent card discovery and task delegation
Fleet Worker Slots
classify
claude-haiku-4-5
Intent classification, fast and cheap
extract
claude-sonnet-4-6
Structured data extraction, JSON output
verify
claude-opus-4-6
Fact checking, validation, highest reasoning
summarize
claude-sonnet-4-6
Content summarization, condensation
generate
claude-opus-4-6
Long-form text and code generation
tool_call
claude-sonnet-4-6
Tool selection and execution, 97.2% success rate