Score a tool call

Inspect an outbound agent tool call, score its risk, apply your policies, and return a decision. Requires the gateway:write scope. When a require_approval policy fires the decision is hold and an approvalId is returned. Poll GET /api/gateway/approvals/{approvalId} for the resolution.

post/api/v1/gateway/tool-call

Request

Base URL: https://app.axiorank.com
URL: https://app.axiorank.com/api/v1/gateway/tool-call
Auth: HTTP bearer

Headers

Idempotency-Keystring

Optional. Safely retry a write: the first response for a given key is replayed on retries (24h TTL). A concurrent retry gets 409; reusing a key with a different body gets 422.

Request body

toolstring required

argumentsobject

phase'request' | 'result' | 'prompt' | 'completion'

resultTextstring

promptTextstring

completionTextstring

contextstring

costUsdnumber

modelstring

inputTokensinteger

outputTokensinteger

durationMsinteger

intentstring

taintHandlestring

passportboolean

source'sdk' | 'mcp' | 'coding-agent' | 'browser-agent'

traceIdstring uuid

stepIndexinteger

sessionIdstring uuid

metadataobject

parentStepIndexinteger

Response

The governance decision.

decision'allow' | 'deny' | 'hold' required

reasonstring required

riskinteger required

auditLogIdstring required

approvalIdstring

Present ONLY when decision is hold. Poll GET /api/gateway/approvals/{id}.

ifcboolean

True when this workspace enforces information-flow control. The SDK uses it to decide whether to report untrusted tool results (phase="result") for taint tracking.

Example response

{
  "signals": [
    {
      "detector": "secret.aws_access_key",
      "location": "body.env.AWS_SECRET"
    }
  ]
}

Changes

No recorded changes to this operation across all 1 revision of this API.