OpenClaw Solana Trader V2 Plugin

Multi-agent Solana memecoin trading plugin — 8-agent analytical team with structured output contracts, cross-agent tasking, and deterministic confidence scoring.

Architecture

The V2 plugin replaces the V1 monolithic single-agent design with a specialized team of 8 agents, each owning a distinct analytical domain. The CTO (Chief Trading Officer) serves as the final decision authority, aggregating scores from specialist agents through a cross-agent tasking protocol.

┌─────────────────────────────────────────────────────────┐
│                     CTO (Agent 1)                       │
│         Final decisions · Portfolio allocation           │
│         Heartbeat: 5m · Full tool access                │
├──────────┬──────────┬───────────┬───────────────────────┤
│          │          │           │                       │
│  On-Chain │  Alpha   │  Social   │  Smart Money         │
│  Analyst  │  Signal  │  Intel    │  Tracker             │
│  Agent 3  │  Agent 4 │  Agent 5  │  Agent 6             │
│           │          │           │                       │
├───────────┴──────────┴───────────┴──────────────────────┤
│                                                         │
│  Execution Specialist (Agent 2)  │  Risk Officer (7)    │
│  Trade timing · Slippage mgmt    │  Portfolio audits     │
│                                  │                       │
│  Strategy Researcher (Agent 8)                          │
│  Weight evolution · Regime detection                    │
└─────────────────────────────────────────────────────────┘

Signal Domains

| Agent | Domain | Access Pattern | |---|---|---| | On-Chain Analyst | Bitquery research, on-chain discovery | On-demand (CTO cross-tasks) + cron | | Alpha Signal Analyst | SpyFly aggregator, external alpha channels | Heartbeat + webhook + cron | | Social Intelligence Analyst | Twitter/X sentiment, narrative detection | On-demand + cron | | Smart Money Tracker | Whale tracking, deployer profiling | On-demand + cron |

Confidence Scoring

V1 Formula: confidence = 0.60 * onchainScore + 0.40 * signalScore - riskPenalty

V2 Formula: Full weighted formula incorporating all 4 signal sources (on-chain, alpha, social, smart money) with self-evolving weights managed by the Strategy Evolution Researcher.

Agent Roster

| ID | Role | Scheduling | Description | |---|---|---|---| | cto | Chief Trading Officer | Heartbeat 5m | Final decision authority. Aggregates analyst scores, manages portfolio. Only agent with state-mutating permissions. | | execution-specialist | Execution Specialist | Heartbeat 3m (V2) | Trade timing, slippage management, exit execution. Does not initiate trades. | | onchain-analyst | On-Chain Analyst | On-demand + cron | Deep Bitquery research, on-chain discovery, deployer profiling. | | alpha-signal-analyst | Alpha Signal Analyst | Heartbeat 5m + webhook | External signal sourcing, caller reputation, signal clustering. | | social-analyst | Social Intelligence | On-demand + cron | Twitter/X trending, sentiment, narrative exhaustion. | | smart-money-tracker | Smart Money Tracker | On-demand + cron | Whale tracking, accumulation patterns, fresh wallet detection. | | risk-officer | Risk & Portfolio Officer | Cron only | Portfolio stress tests, exposure checks. Advisory — does not execute. | | strategy-researcher | Strategy Researcher | Cron only | Trade outcome analysis, weight adjustments, regime detection. |

ID mapping note: Gateway configs and openclaw.plugin.json use canonical agent IDs listed above. The skill folder social-intel-analyst/ maps to agent ID social-analyst. Architecture docs may reference shortened aliases (e.g., alpha-analyst for alpha-signal-analyst) — always use the full canonical ID in configs.

Setup

1. Plugin Registration

Register the plugin with your OpenClaw Gateway using openclaw.plugin.json. The plugin ID is solana-trader-v2.

2. Configuration Fields

| Field | Required | Description | |---|---|---| | orchestratorUrl | Yes | Base URL of the trading orchestrator API | | apiKey | Yes | API key from signup or https://traderclaw.ai/register | | walletId | No | Wallet UUID (auto-created on first session if omitted) | | gatewayBaseUrl | No | Public HTTPS URL of your Gateway for event-to-agent forwarding | | gatewayToken | No | Bearer token for authenticating forwarded events | | activeAgents | No | Agent IDs to activate. Defaults to all 8 if omitted. | | agentId | No | Primary agent ID for subscription event routing | | apiTimeout | No | HTTP timeout in ms (default: 30000) |

3. Gateway Config

Copy the appropriate gateway config to your Gateway's configuration directory:

• V1 rollout (5 agents): config/gateway-v1.json5
• V2 full team (8 agents): config/gateway-v2.json5

Replace "shared-secret" in the hooks.token field with your actual webhook authentication token.

4. Webhook Endpoints

Configure your upstream services to deliver events to:

| Endpoint | Source | Target Agent | |---|---|---| | /hooks/alpha-signal | SpyFly alpha aggregator | alpha-signal-analyst | | /hooks/firehose-alert | Bitquery event-to-agent forwarding | onchain-analyst |

Scheduling

The plugin uses a dual-loop scheduling architecture:

Fast Loop (Heartbeat)

Runs every 3-5 minutes depending on agent. Handles real-time trading: scanning, analysis, decision-making, execution, and position monitoring.

Slow Loop (Cron)

Runs on hour/multi-hour cadences in isolated sessions. Handles retrospective work that benefits from data accumulation.

| Cron Job | Schedule | Agent | Purpose | |---|---|---|---| | strategy-evolution | Every 4h | strategy-researcher | Weight adjustments from trade outcomes | | risk-audit | Every 2h | risk-officer | Portfolio stress tests, exposure checks | | source-reputation | Every 3h | alpha-signal-analyst | Alpha source win/loss analysis | | meta-rotation | Every 3h (:30) | onchain-analyst | Narrative cluster analysis | | whale-watch | Every 2h | smart-money-tracker (V2) / onchain-analyst (V1) | Large wallet movement scanning | | sentiment-trend | Every 4h | social-analyst (V2 only) | Twitter/X trend analysis | | dead-money-sweep | Every 2h | cto | Exit stale positions | | subscription-cleanup | Every 1h | cto | Prune unused Bitquery subscriptions | | daily-report | Daily 04:00 UTC | cto | PnL summary, performance metrics | | execution-health | Every 6h | execution-specialist (V2 only) | Slippage and execution quality review |

Cron jobs use CRON_JOB: message prefix. Agents detect this prefix and skip the full trading loop, executing only the specified job.

V1 vs V2 Differences

| Aspect | V1 | V2 | |---|---|---| | Active agents | 5 (CTO, On-Chain, Alpha, Risk, Strategy) | All 8 | | Disabled agents | Execution Specialist, Social, Smart Money | None | | Cron jobs | 8 | 10 | | Max concurrent cron runs | 2 | 3 | | Whale tracking | Consolidated into onchain-analyst | Dedicated smart-money-tracker | | Sentiment analysis | Not available | social-analyst with cron job | | Execution monitoring | CTO handles directly | Dedicated execution-specialist (3m heartbeat) | | Confidence formula | 2-source (on-chain + alpha) | 4-source (on-chain + alpha + social + smart money) |

Upgrade Path (V1 → V2)

1. Replace config/gateway-v1.json5 with config/gateway-v2.json5
2. Ensure all 8 agent skills are deployed to the Gateway
3. Update activeAgents in plugin config (or remove to default to all 8)
4. The whale-watch cron job automatically moves from onchain-analyst to smart-money-tracker
5. New cron jobs (sentiment-trend, execution-health) activate automatically
6. maxConcurrentRuns increases from 2 → 3 to handle additional cron load

Directory Structure

openclaw-plugin-v2/
├── openclaw.plugin.json          # Plugin manifest (agent IDs, config schema)
├── package.json                  # Package definition
├── index.ts                      # Plugin entry point
├── config/
│   ├── gateway-v1.json5          # V1 Gateway config (5 agents, 8 cron jobs)
│   └── gateway-v2.json5          # V2 Gateway config (8 agents, 10 cron jobs)
├── skills/                       # Per-agent SKILL.md files
│   ├── alpha-signal-analyst/
│   ├── execution-specialist/
│   ├── onchain-analyst/
│   ├── risk-officer/
│   ├── smart-money-tracker/
│   ├── social-intel-analyst/
│   └── strategy-researcher/
└── src/
    ├── agent-registry.ts         # Canonical agent definitions, tool permissions
    ├── cross-tasking.ts          # Cross-agent message bus, permission matrix
    ├── cross-tasking.test.ts     # 90 tests for cross-tasking
    ├── shared-context.ts         # Shared context payload builder
    ├── output-contracts.ts       # Structured output schemas per agent
    ├── output-validator.ts       # Runtime output validation
    ├── confidence.ts             # Deterministic confidence scoring
    ├── session-manager.ts        # Agent session lifecycle
    ├── http-client.ts            # Orchestrator HTTP client
    ├── alpha-buffer.ts           # In-memory alpha signal ring buffer
    └── alpha-ws.ts               # Alpha WebSocket manager