Structured Protocols

Give Your AI Agents Real Contracts

Stop sending vague instructions. Define what agents must deliver, what they receive, and how their output feeds into the next agent.

Vague Tasks Produce Vague Results

Most AI agent workflows start with a text description: "Add authentication to the API." The agent interprets this however it wants, and you hope the output matches your expectations.

This works for simple tasks. It breaks down when:

Multiple agents need to produce compatible output
Downstream agents depend on specific artifacts from upstream agents
You need to verify that an agent actually delivered what was requested
Complex tasks require distinguishing must-have from nice-to-have requirements

Without structured specs, every task is a roll of the dice.

Structured Task Specs & Contract Exchange

Orchestratia introduces a formal protocol for task specifications and result reporting.

graph TB
    Spec["Structured Spec<br/>MoSCoW + Constraints + Output Schema"]
    Agent["AI Agent<br/>Executes Task"]
    Result["Structured Result<br/>Contract Fulfillment Report"]
    Next["Downstream Agent<br/>Receives resolved_inputs"]

    Spec --> Agent
    Agent --> Result
    Result -->|"Contract exchange"| Next

    style Spec fill:#F0F9F7,stroke:#2A9D88,stroke-width:2px
    style Agent fill:#FAF8F5,stroke:#E8E3DA
    style Result fill:#F0F9F7,stroke:#2A9D88,stroke-width:2px
    style Next fill:#FAF8F5,stroke:#E8E3DA

Task Spec Format

Each task can include a structured_spec with:

MoSCoW requirements — Must Have, Should Have, Could Have, Won't Have
Constraints — Technology choices, patterns to follow, files to avoid
Output expectations — What the agent must produce (API schema, test results, migration file)
Acceptance criteria — Concrete, verifiable conditions for completion