Connector docs
The exact access every connector asks for
No adjectives, the policy. For each connector: the literal API call CostMon makes, the minimum permission it needs, the fields you supply, what it can't see, and every caveat. No login required to read this.
Cloud
AWS
CostMon calls AWS Cost Explorer's GetCostAndUsage once per sync, daily granularity, grouped by linked account and service. Nothing else in your AWS account is touched.
Read the access reference →GCP
CostMon runs a read-only BigQuery SQL query against the FOCUS-formatted billing export table you point it at. It never calls any other GCP API.
Read the access reference →Azure
CostMon calls Azure Cost Management's Query API once per sync, daily granularity, grouped by service name and region, scoped to one subscription.
Read the access reference →AI
Anthropic
CostMon calls two read-only Anthropic Admin API endpoints: the organization cost report, then the messages usage report to add a token breakdown.
Read the access reference →OpenAI
CostMon calls two OpenAI organization-level reporting endpoints and merges them per day and per model. Neither is reachable with a regular project API key.
Read the access reference →Amazon Bedrock
CostMon joins AWS Cost Explorer's Bedrock spend with CloudWatch's per-model token metrics, using the same read-only, cost-scoped pattern as the AWS connector.
Read the access reference →Vertex AI / Gemini
CostMon reads the same FOCUS-formatted GCP billing export as the gcp connector, filtered to Vertex AI / Generative Language services, and joins it with Cloud Monitoring's per-model token metrics.
Read the access reference →Helicone
CostMon queries Helicone's gateway request log and aggregates it to a daily grain by model, upstream provider, and workload. It never stores one row per request.
Read the access reference →Data & SaaS
Snowflake
CostMon runs a read-only SQL statement against Snowflake's own organization usage view, via the SQL API, using a role scoped to metering data only.
Read the access reference →Databricks
CostMon runs a read-only SQL statement against system.billing.usage joined to system.billing.list_prices via the Databricks SQL Statement Execution API.
Read the access reference →Datadog
CostMon calls Datadog's Usage Metering API: estimated_cost for the current and previous month, historical_cost for anything older the sync window reaches.
Read the access reference →GitHub
CostMon calls GitHub's billing usage report for either an organization or an enterprise, scoped to the read-only Administration permission.
Read the access reference →Vercel
CostMon calls Vercel's FOCUS v1.3 billing charges API and maps the newline-delimited JSON records it returns straight into normalized cost rows.
Read the access reference →Bring any vendor
No connector yet? Push the numbers directly.
POST /v1/costs takes normalized cost rows from anything, and a CSV import path covers the rest. Same idempotency guarantee as every built-in connector: re-sending a window replaces amounts instead of duplicating rows.
A rate you looked up will change again.
Provider pricing pages change without notice. CostMon tracks the bill that pays for it.