Skip to main content

Vantage Integration

Connect Vantage so Autoheal can query your cost data and root-cause anomalies across your cloud providers. When a Vantage cost or anomaly alert fires, an agent pulls the underlying cost time series. It then pivots to AWS / GCP / Kubernetes tools to identify the workload responsible for the spike.

Capabilities​

CapabilityDescription
Query Cost DataPull cost time series for any CostReport or VQL filter
List Anomalies & RecommendationsSurface active anomalies and cost-saving recommendations from Vantage
Cross-Provider RCACombine Vantage data with AWS / GCP / Kubernetes tools to locate the workload driving the spend
Trigger Investigations from Vantage AlertsRelay Vantage cost / anomaly alerts into Autoheal via Jira or Slack (see below)

Prerequisites​

  • A Vantage account with API access
  • A Vantage API token (a read-only token is sufficient)

Setup​

1
Generate a Vantage API Token
  1. In Vantage, go to Settings → API Tokens
  2. Create a new token (read-only is fine) and copy the value
note
Vantage API tokens are used directly as Authorization: Bearer credentials. No OAuth client-credentials exchange is performed, despite what the public OAS suggests.
2
Add the Integration in Autoheal
  1. Go to Integrations in Autoheal
  2. Click Vantage
  3. Enter a name (e.g., "Production Vantage")
  4. Paste your API Token
  5. Click Save

Configuration Fields​

FieldRequiredDescription
API TokenYesVantage API token from Settings → API Tokens. Read-only is sufficient.

Triggering Investigations on Vantage Alerts​

Vantage's Cost Alerts and Anomaly Alerts only deliver notifications to Email, Slack / Microsoft Teams, and Jira. There's no native webhook destination. To get Autoheal to auto-investigate when Vantage fires an alert, use one of these relays:

If you already use the Jira integration with Autoheal:

  1. In Vantage, open the Cost Alert (or Anomaly Alert) you want Autoheal to investigate
  2. In Setup App Alerts, add a Jira Alerts destination. Point it at the same Jira project that's connected to Autoheal
  3. Include the anomaly token in the Jira ticket description (e.g., Vantage anomaly: anml_AbCdEfGh1234). Agents can then call vantage_get_anomaly_alert for the anomaly detail
  4. When Vantage fires, it creates a Jira ticket → Jira webhook hits Autoheal → Autoheal opens an investigation. The investigation body contains the Vantage anomaly token

Agents recognise the Vantage anomaly token in the ticket body. They pivot directly into Vantage MCP tools for the cost detail, then into cross-platform tools (gcloud / aws) for resource-level RCA.

Slack mention​

Have Vantage post the alert to Slack, then @autoheal mention the message in-thread:

@autoheal investigate this Vantage cost anomaly

Manual / agent-on-demand​

You can always trigger an investigation by asking an agent directly:

Investigate the Vantage anomaly anml_AbCdEfGh1234
What caused our AWS EC2 spike yesterday? Check Vantage and AWS.

MCP Tools​

When agents have access to this integration, they can call:

ToolWhat it does
vantage_list_anomaly_alertsList cost anomaly alerts with filters (provider, service, status, date range)
vantage_get_anomaly_alertFull detail for a single anomaly — amount, baseline, resources, resource_tokens
vantage_list_cost_reportsList CostReports in the workspace
vantage_get_costsPull cost time series for a CostReport or VQL filter, with date_bin + groupings
vantage_list_recommendationsList cost-saving recommendations (rightsizing, idle, commitment, …)
vantage_list_workspacesList workspaces the token can access
vantage_list_cost_providersList cloud providers visible in a workspace
vantage_list_cost_servicesList services per provider (for VQL filter discovery)
vantage_test_connectionVerify the API token (debug tool)

vantage_get_costs is response-size aware: it defaults to a 30-day window, day-level binning, and grouping by service. If a response would exceed the soft size cap, the tool refuses with a structured error. It does not silently return a partial time series. The error tells the agent to narrow the window, coarsen date_bin, or drop groupings.

How It Works​

  1. A Vantage Cost / Anomaly Alert fires and lands in your Slack or Jira (via Vantage's native delivery)
  2. The Slack mention or Jira-webhook relay opens an Autoheal investigation
  3. The agent calls vantage_get_anomaly_alert for full detail, then vantage_get_costs on the originating CostReport to confirm the spike timing
  4. Using the resources list, the agent pivots to AWS / GCP tools to identify the workload behind the spend
  5. Findings land on the investigation. You see the root cause, and you do not need to search Vantage manually
tip

Add an integration guide skill that lists your most-watched CostReports and provider/service tuples. This lets agents skip discovery and go straight to root-causing.

Example Queries​

Show me the active cost anomalies from the last 7 days
Why did our AWS EC2 spend spike yesterday?
What are the largest cost-saving recommendations Vantage has for our GCP account?

Troubleshooting​

vantage_get_costs Returns _truncated

Cost time-series responses can dwarf an agent's context window. When you see _truncated:

  • Narrow start_date / end_date
  • Coarsen date_bin (use month for windows over 60 days)
  • Reduce the number of groupings
API Returns 401 / 403
  • Check that the API Token is current (not revoked) and has access to the workspaces you expect
  • If the token has access to multiple workspaces, several tools require you to pass workspace_token. Check the tool's argument schema
Agent Can't Find the Anomaly from a Jira Relay

Make sure the Vantage → Jira alert template includes the anomaly token (e.g., anml_AbCdEfGh1234) in the ticket title or description. Without the token, agents can't call vantage_get_anomaly_alert directly and must use vantage_list_anomaly_alerts instead.