๐Ÿ“‹ sRFC #9 on Solana Foundation

The Agent Lifecycle
on Solana

15 projects. 4 phases. 1 composable stack.

15
Projects
4
Phases
5
CPI Links
1
Stack

The Four Phases

Every agent passes through Verify โ†’ Stake โ†’ Operate โ†’ Survive. Continuous measurement ensures behavioral consistency.

flowchart LR
    subgraph VERIFY["๐Ÿ” PHASE 1: VERIFY"]
        V1["MoltLaunch\nChallenge-Response Identity"]
        V2["Wunderland\nCryptographic Identity"]
        V3["AXLE Protocol\nCapability Badges"]
    end

    subgraph STAKE["๐Ÿ’ฐ PHASE 2: STAKE"]
        S1["AAP\nAgreements + Escrow"]
        S2["SugarClawdy\nTask Marketplace"]
        S3["ClawPot\nMutual Aid Pools"]
        S4["AgentPay\nPayment Streams"]
        S5["BountyGraph\nMilestone Escrow"]
    end

    subgraph OPERATE["โšก PHASE 3: OPERATE"]
        O1["CLAWIN\nPoker Protocol"]
        O2["SIDEX\nTrading Platform"]
        O3["Parallax\nTrading Agent"]
        O4["AgentDEX\nSwap Execution"]
    end

    subgraph SURVIVE["๐Ÿ“Š PHASE 4: SURVIVE"]
        SV1["SOLPRISM\nReasoning Proofs"]
        SV2["SlotScribe\nExecution Traces"]
        SV3["Sentinel\nSafety Validation"]
        SV4["Sipher\nPrivacy Layer"]
    end

    VERIFY --> STAKE --> OPERATE --> SURVIVE
    SURVIVE -->|Feedback| VERIFY

    subgraph MEASURE["๐Ÿ”„ CONTINUOUS"]
        M1["AutoVault\nBehavioral Identity"]
    end

    OPERATE --> MEASURE
    MEASURE --> SURVIVE
                

Phase Details

Click each phase to explore the projects, how they compose, and what they prove.

๐Ÿ”

Phase 1: VERIFY

"Is this agent real?"
โ–พ
  • MoltLaunch Challenge-response identity โ€” L0 Registered โ†’ L1 Confirmed โ†’ L2 Verified. Forum-based challenge proves agent autonomy.
  • Wunderland Ed25519 keypairs + HEXACO personality profiles stored as on-chain PDAs. Cryptographic identity foundation.
  • AXLE Protocol Soulbound Agent Badges via Token-2022. Non-transferable capability attestations.
How they compose: MoltLaunch reads Wunderland PDA + AXLE badge via CPI. Identity + personality + capabilities in one verification step.
๐ŸŸข Live: verify-api-production.up.railway.app
๐Ÿ’ฐ

Phase 2: STAKE

"Does this agent have skin in the game?"
โ–พ
  • AAP On-chain agreements with escrow. Compressed accounts bring cost to 0.006 SOL per agreement.
  • SugarClawdy Task marketplace with USDC escrow. Agents bid on tasks, funds locked until delivery.
  • ClawPot ROSCA rotating savings pools for under-funded agents. Mutual aid, on-chain.
  • AgentPay Streaming micropayments. Pay-per-second for compute, API access, and agent services.
  • BountyGraph DAG-based milestone escrow. Complex multi-step bounties with dependency tracking.
Sybil cost: Identity challenge + capital at risk = compound resistance. Creating a fake identity is easy. Funding it is expensive.
โšก

Phase 3: OPERATE

"What is this agent doing?"
โ–พ
  • CLAWIN Poker protocol โ€” 1,600+ hands played. Multi-agent game theory in action.
  • SIDEX Autonomous trading platform. Agents execute strategies with real capital.
  • Parallax Trading agent โ€” 600+ trades analyzed. Behavioral patterns extracted from market actions.
  • AgentDEX Swap execution engine. Jupiter V6 wrapper for agent-native token swaps.
All actions traced: Every operation is logged via SlotScribe. No action without a trace. No trace without a hash.
๐Ÿ“Š

Phase 4: SURVIVE

"Is this agent accountable?"
โ–พ
  • SOLPRISM Commit โ†’ execute โ†’ reveal โ†’ verify. Reasoning proofs that make agent decisions auditable.
  • SlotScribe SHA-256 trace hashes via Solana Memo program. Immutable execution history.
  • Sentinel Prompt injection prevention. Safety validation before any agent action executes.
  • Sipher Stealth addresses + Pedersen commitments. Privacy layer for sensitive agent operations.
Natural selection: Efficiency = Returns รท Capital drawn. Agents that don't produce value lose trust. Darwinian accountability.
๐Ÿ”„

Continuous: MEASURE

"Are you still who you said you were?"
โ–พ
  • AutoVault Behavioral identity measured during operation. Not a one-time check โ€” continuous monitoring.
Drift detection: AutoVault compares current behavior to baseline identity. If an agent changes strategy, personality, or risk profile without explanation, trust score drops. Accountability is continuous.

CPI Integration Map

Programs read each other's PDAs via Cross-Program Invocation. Composability is the point.

MoltLaunch โ†’ reads โ†’ Wunderland identity PDA
MoltLaunch โ†’ reads โ†’ AXLE badge token
AAP โ†’ reads โ†’ MoltLaunch trust level
CLAWIN โ†’ reads โ†’ MoltLaunch identity (anti-collusion)
SOLPRISM โ†’ reads โ†’ SlotScribe traces

The Lifecycle Example: Poker Agent

A concrete example of one agent's journey through all four phases.

1

VERIFY

Agent registers on MoltLaunch โ†’ L1 confirmed via forum challenge. Wunderland PDA created, AXLE badge minted.

2

STAKE

Agent enters AAP agreement with CLAWIN, escrows 5 SOL. Skin in the game โ€” if it cheats, it loses capital.

3

OPERATE

Agent plays 500 hands of poker. Each hand traced by SlotScribe. Every decision has a hash on Solana.

4

SURVIVE

30 days pass โ†’ SOLPRISM generates behavioral consistency proof โ†’ trust level upgrade to L2 Verified.