Pilot Use Cases

Three concrete scenarios for SwarmOS v1.0.0-rc1 pilots. Each uses the local chain adapter with capped funds.

1. DeFi Yield Agent

Scenario: An autonomous agent executes yield farming strategies across protocols.

• 1 operator, 3 agents (one per strategy)
• Conservative policy: 100/action limit, daily cap 1000
• Each agent gets its own vault with strategy-specific policy
• HIGH+ risk actions require operator approval

What SwarmOS provides: Per-strategy spend isolation, automatic risk scoring, correlated failure detection via GSI, and hash-chained receipts for strategy auditing.

2. Payment Processing Bot

Scenario: An AI agent processes customer payments with human approval for large amounts.

• 1 operator, 1 payment agent
• Standard policy: 500/action, daily cap 5000
• Single vault with operating funds

What SwarmOS provides: Per-transaction policy enforcement, approval workflow for large amounts, session key rotation, and SAFE_READONLY on signature failure spikes.

3. Multi-Agent Treasury Management

Scenario: A team runs 10-20 agents across research, trading, and operations — each with different risk profiles.

• 2 operators, 15 agents across 3 tiers
• Tier 1 (research): permissive, 2000/action
• Tier 2 (trading): standard, 500/action
• Tier 3 (ops): conservative, 100/action

What SwarmOS provides: Role-based rate limiting, tiered policy templates, cross-agent anomaly detection, and emergency controls requiring multi-operator consensus for HARD_STOP release.

Pilot Parameters

For all pilots:

• Funds: Capped at test amounts (local adapter, no real chain)
• Duration: 5-7 days
• Mandatory drills: SAFE_READONLY, stale session, recovery replay, verifier status
• Success criteria: Zero policy breaches, zero cross-cell contamination, operators can explain failure modes