AI analysis

Overview

The AI analysis endpoints provide Claude-powered change impact analysis. When you have virtual changes staged against a connection, the AI engine analyses them against the full configuration snapshot, returning a risk assessment, summary, and actionable recommendations.

All analysis endpoints require READ_ONLY permission or above. POST is used because of request body size, not because the endpoints modify data.

Analyse virtual changes

POST /v1/analysis/ai

Runs an AI-powered impact analysis on the virtual changes currently staged for the specified connection. You may optionally supply a snapshotId to pin the analysis to a specific configuration snapshot.

curl -X POST https://api.configly.app/v1/analysis/ai \
  -H "Authorization: Bearer cly_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "connectionId": "your_connection_id"
  }'

The response contains a risk assessment, a human-readable summary, and a list of recommendations.

Usage reporting

GET /v1/analysis/ai/usage

Returns aggregated AI usage statistics for the connection scoped to your API key. Use the days query parameter to set the reporting window (default 30, maximum 365).

curl https://api.configly.app/v1/analysis/ai/usage?days=90 \
  -H "Authorization: Bearer cly_your_key"

{
  "data": {
    "windowDays": 90,
    "totalCalls": 47,
    "totalInputTokens": 128540,
    "totalOutputTokens": 31200,
    "totalCostUsd": 1.23
  }
}

Error codes

For deterministic impact analysis, see Impact analysis.