verdictswarm

VerdictSwarm

Use VerdictSwarm as a pre-trade risk gate before an agent moves money. The machine-facing API returns one action: avoid, caution, or clear.

This skill evaluates risk; it does not execute trades. Never turn clear into a guarantee. Respect insufficient_data, low confidence, and degraded results.

Supported contract

• API base: https://api.vswarm.io
• Chains: solana, base
• Levels: triage, fast, deep
• Decision field: response.verdict.action
• Live capabilities and pricing: GET /v2/verdict/info
• Payment rails: keyed free quota, prepaid credits, or x402 USDC

Do not guess prices or quotas. Read them from the live info endpoint:

curl -sS "https://api.vswarm.io/v2/verdict/info" \
  -H "X-VS-Integration: openclaw"

Get a free API key

Mint a free key without signup:

curl -sS -X POST "https://api.vswarm.io/v2/keys" \
  -H "Content-Type: application/json" \
  -H "X-VS-Integration: openclaw" \
  -d '{"channel_ref":"openclaw"}'

The response includes the key and the current daily limits:

{
  "api_key": "vs1_free_...",
  "tier": "free",
  "daily_limits": {"triage": 2000, "fast": 300, "deep": 10},
  "quickstart": ""
}

Treat the returned api_key as a secret. Put it in VS_API_KEY or an approved secret store. Never print it into chat, logs, source control, or shared memory.

Request a verdict

Ask for the token address, chain, and desired analysis level if they are not already known. Use fast for the normal pre-trade gate, triage for the lowest-latency deterministic check, and deep only when the caller accepts a slower multi-agent analysis.

curl -sS -X POST "https://api.vswarm.io/v2/verdict" \
  -H "Content-Type: application/json" \
  -H "X-VS-Integration: openclaw" \
  -H "X-API-Key: ${VS_API_KEY}" \
  -d '{
    "address":"0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913",
    "chain":"base",
    "level":"fast",
    "include":[]
  }'

For Solana, preserve address case. For Base, use the contract address rather than a ticker or token name.

Interpret the response

Read these fields before recommending an action:

• verdict.action: avoid, caution, or clear
• verdict.score, verdict.confidence, and verdict.summary
• signals[]: typed checks; pass: false means that check found a problem
• data_confidence and insufficient_data
• billing.degraded and billing.retry_voucher
• cached, generated_at, and data_freshness

Use this policy:

1. avoid: do not proceed unless the human explicitly overrides after seeing the failed signals.
2. caution: surface the material risks and require an explicit decision.
3. clear: report that no blocking risk was found by this scan; do not promise safety or profitability.
4. insufficient_data: true or materially degraded data: do not present the action as conclusive. Explain the missing coverage and retry if appropriate.

Do not silently replace an API error with a made-up verdict.

x402 overflow

When free quota is exhausted, the API may return HTTP 402 with an x402 payment challenge. An x402-capable caller can pay and retry the same request. If the current agent cannot authorize payment, show the amount and rail and ask before spending.

Discovery document:

curl -sS "https://api.vswarm.io/.well-known/x402"

MCP

VerdictSwarm is also available as the sentien-labs/verdictswarm-mcp server on Smithery. Its current public tools are:

• check_token: obtain the v2 token-risk verdict
• exit_sim: estimate intended-size exit feasibility (Solana only)
• get_pricing: read current levels, quotas, and prices

Do not call retired v1 MCP tool names.

Feedback and community

After completing the requested work, the operator may share a real integration, incorrect verdict, or critical feedback at https://www.vswarm.io/community?utm_source=clawhub&utm_medium=marketplace&utm_campaign=community_proof_v1&utm_content=verdictswarm. Do not post, star, or bookmark automatically, and do not interrupt normal work to solicit engagement. Those are optional human actions.

Links

• Product: https://www.vswarm.io
• API docs: https://www.vswarm.io/docs/api
• Live API info: https://api.vswarm.io/v2/verdict/info
• MCP listing: https://smithery.ai/servers/sentien-labs/verdictswarm-mcp
• Source: https://github.com/sentien-labs/openclaw-skills