Coordination Modes

YorN supports five coordination modes that define how multiple agents work together to make trading decisions. Each mode represents a different architecture for combining agent signals into a unified trading action.

Overview

When you deploy multiple agents as a team, the coordination mode determines how their individual signals are aggregated, filtered, and acted upon. Choosing the right mode depends on your goals: speed, consensus, specialization, depth of analysis, or behaviorally varied crowd simulation.

Independent Agents

Without coordination, each agent trades independently. Good for uncorrelated strategies, but risks position conflicts.

Coordinated Teams

With coordination, agents share a unified view of the portfolio and combine their signals through the selected mode.

Swarm Mode

All agents vote simultaneously. The majority signal wins.

Latency

Low

Consensus

Medium

Best For

Fast Markets

In Swarm mode, all agents in the team evaluate the same market data concurrently. Each agent produces a Signal independently. The coordinator tallies all signals and executes the majority decision, weighted by confidence. If agents are split, the trade is skipped.

                Swarm Configuration
                JSON
            

{
    "team_name": "NBA-Swarm",
    "coordination_mode": "swarm",
    "agents": ["agent_001", "agent_002", "agent_003"],
    "swarm_config": {
        "quorum": 0.6,              // 60% must agree
        "weight_by_confidence": true,
        "timeout_ms": 5000,          // Max wait for all votes
        "tie_action": "skip"          // "skip" | "defer_to_lead"
    }
}
            

Council Mode

A lead agent proposes, others approve or veto.

Latency

Medium

Consensus

High

Best For

High-Stake Trades

In Council mode, one agent is designated as the lead. The lead proposes a trade, and the remaining agents act as reviewers. Each reviewer can approve, abstain, or veto. A veto from any reviewer blocks the trade. This mode is ideal for high-stakes markets where false positives are costly.

                Council Configuration
                JSON
            

{
    "team_name": "FOMC-Council",
    "coordination_mode": "council",
    "agents": ["analyst_lead", "risk_reviewer", "macro_reviewer"],
    "council_config": {
        "lead_agent": "analyst_lead",
        "veto_enabled": true,
        "approval_threshold": 0.5,  // 50% of reviewers must approve
        "review_timeout_ms": 10000,
        "abstain_counts_as": "approve"  // "approve" | "skip"
    }
}
            

Pipeline Mode

Agents run in sequence, each refining the previous signal.

Latency

High

Accuracy

High

Best For

Complex Analysis

In Pipeline mode, agents execute in a defined sequence. The first agent generates an initial signal, which is passed as enriched context to the second agent, and so on. Each stage can refine the confidence, adjust the position size, or cancel the trade entirely. This mode is best when different agents specialize in different aspects of analysis (e.g., data collection, signal generation, risk assessment).

                Pipeline Configuration
                JSON
            

{
    "team_name": "Weather-Pipeline",
    "coordination_mode": "pipeline",
    "pipeline_config": {
        "stages": [
            {
                "agent": "data_collector",
                "role": "Gather ensemble forecast data",
                "timeout_ms": 15000
            },
            {
                "agent": "signal_generator",
                "role": "Generate buy/sell signal from forecast",
                "timeout_ms": 10000
            },
            {
                "agent": "risk_gate",
                "role": "Final risk check and position sizing",
                "timeout_ms": 5000
            }
        ],
        "abort_on_null": true,      // Stop pipeline if any stage returns None
        "pass_context": true        // Each stage receives previous output
    }
}
            

Deliberation Mode

Agents debate in rounds until consensus is reached.

Latency

Very High

Consensus

Very High

Best For

Novel Markets

In Deliberation mode, agents engage in structured rounds of debate. Each agent presents its signal and reasoning. In subsequent rounds, agents can revise their signals based on arguments from other agents. The process continues until consensus is reached or the maximum number of rounds is exhausted. This mode uses Claude as the deliberation mediator to synthesize arguments.

                Deliberation Configuration
                JSON
            

{
    "team_name": "Politics-Debate",
    "coordination_mode": "deliberation",
    "agents": ["poll_analyst", "sentiment_reader", "contrarian"],
    "deliberation_config": {
        "max_rounds": 3,
        "consensus_threshold": 0.8,  // 80% agreement to act
        "mediator": "claude",         // AI mediator for synthesis
        "round_timeout_ms": 15000,
        "require_reasoning": true,    // Agents must explain signals
        "no_consensus_action": "skip"  // "skip" | "majority"
    }
}
            

Deliberation mode uses additional Claude API calls for mediation. This counts against your daily AI usage quota. Enterprise plans have higher limits for deliberation-heavy workflows.

Mode Comparison

Use this table to select the right coordination mode for your use case:

Property	Swarm	Council	Pipeline	Deliberation
Latency	Low	Medium	High	Very High
Consensus Quality	Medium	High	High	Very High
Best For	Fast-moving markets	High-stakes trades	Multi-step analysis	Novel or uncertain markets
Min Agents	3	2	2	2
Max Agents	Unlimited	10	8	5
AI Usage	Low	Low	Medium	High
Conflict Resolution	Majority vote	Veto power	Sequential override	AI-mediated debate
Tier Required	Pro	Pro	Pro	Enterprise

Configuration

Teams are configured via the API or the dashboard Team Builder interface. Here is a complete example that creates a team using the API:

                Create Coordinated Team
                POST /api/teams/create
            

POST /api/teams/create
Authorization: Bearer <token>
Content-Type: application/json

{
    "name": "Economics-Swarm-v1",
    "coordination_mode": "swarm",
    "market_category": "economics",
    "agents": [
        {
            "template": "fomc_hold",
            "name": "FOMC-Analyst",
            "params": { "futures_threshold": 0.75 }
        },
        {
            "template": "jobs_surprise",
            "name": "Jobs-Tracker",
            "params": { "adp_weight": 0.4 }
        },
        {
            "prompt": "Contrarian: buy NO when consensus is >85% YES",
            "name": "Contrarian"
        }
    ],
    "swarm_config": {
        "quorum": 0.66,
        "weight_by_confidence": true
    },
    "mode": "paper",
    "max_position": 75,
    "daily_loss_limit": 100
}
            

Coordination modes require a Pro or Enterprise tier subscription. Free tier accounts are limited to independent agent operation.

Next Steps

Agent Creation -- Build the individual agents for your team
Strategies -- Create custom AI-generated strategies
API Reference -- Full team management endpoints