Skip to main content
Private access is opening up — request an invite
askFinz
Protocols · Orchestration

The askFinz Orchestrator.

A coordination layer that routes work between 50+ specialized agents using open protocols. Built for interoperability — agents can talk to each other, to you, to external tools, and to the UI in structured ways.

How it works

Every prompt flows through the orchestrator.

The orchestrator is a small set of foundational agents that sit between you and the specialist agents. It reads intent, plans steps, routes work, and checks quality before results reach you.

Intent Classification

The Thinker agent reads your prompt and classifies intent, complexity, and required capabilities before routing.

Multi-Step Planning

The Planner breaks complex tasks into ordered steps, each tagged with the specialist agent needed.

Dynamic Routing

The Router matches each step to the best-fit agent based on capabilities, current load, and reasoning depth required.

Parallel Execution

Independent steps run in parallel. The Executor coordinates timing and merges results when dependencies resolve.

Quality Review

The Reflector checks finished work against the original brief and requests retries when quality falls short.

Memory Integration

The Ingestor pulls relevant context from prior conversations, uploaded files, and saved work before execution.

Open protocols

Built on standards, not silos.

askFinz agents communicate using open protocols. This means external agents can call askFinz, askFinz agents can call external tools, and you can inspect the messages flowing between them.

A2A (Agent-to-Agent)

Shipping

Agents communicate directly with each other to coordinate complex tasks. The askFinz Orchestrator routes work between specialized agents using structured messages.

Use cases: Internal agent coordination, sub-agent spawning, parallel task execution

A2P (Agent-to-Person)

Soon

Agents pause for human input when needed. Permission requests, confirmations, and clarifications flow through this protocol.

Use cases: Permission requests, CAPTCHA pauses, decision points, review gates

MCP (Model Context Protocol)

Shipping

Anthropic's open standard for connecting AI systems to external data sources and tools. askFinz agents can call MCP servers you configure.

Use cases: External tool integration, custom data sources, third-party services

A2UI (Agent-to-UI)

Soon

Agents stream structured updates to the interface. Progress indicators, intermediate results, and tool calls render in real-time.

Use cases: Live progress streaming, step-by-step visibility, result rendering
Why this matters

Interoperability by design.

When agents speak open protocols, you're not locked into one vendor's ecosystem. External agents can call askFinz. askFinz agents can call your tools. The messages are structured and inspectable.

The orchestrator coordinates this without becoming a bottleneck. Agents hand work to each other directly when they can. The orchestrator steps in only when routing decisions or quality checks are needed.

This is how we keep the platform open at the edges while maintaining quality and safety in the middle.

Want to see the orchestrator in action? The agent flow diagram shows how work moves through the system.