Statsig Dashboard

Create Statsig Dashboard

Build valid dashboard API requests for the console v1 dashboards API and execute them with the bundled scripts.

Read references/dashboard-api.md for the exact field-level schema, supported values, and ready-to-edit examples.

The command examples below assume you are running from this skill directory.

Quick Start

1. Export the API key:

export STATSIG_CONSOLE_API_KEY="..."

1. Save the request body to a JSON file.
2. Run the matching script:

python3 scripts/create_dashboard.py \
  --body-file /tmp/dashboard.json

Companion Scripts

• scripts/create_dashboard.py: create a new dashboard with POST /console/v1/dashboards
• scripts/read_dashboard.py: fetch a dashboard and print a create-compatible shape by default using GET /console/v1/dashboards/{id}
• scripts/add_dashboard_widgets.py: append widgets with POST /console/v1/dashboards/{id}/widgets
• scripts/replace_dashboard_widgets.py: replace all widgets with PUT /console/v1/dashboards/{id}/widgets

Workflow

1. Choose the correct endpoint:

• create a dashboard: create_dashboard.py
• read a dashboard in create-compatible shape: read_dashboard.py
• append widgets: add_dashboard_widgets.py
• replace all widgets: replace_dashboard_widgets.py

1. Build the matching request body for the chosen endpoint.
2. Include only supported fields:

• name
• description
• defaults
• widgets

1. Map each requested widget to one of the supported widget types:

• header
• text
• timeseries
• each widget may include optional width and height

1. For timeseries, enforce the query contract exactly:

• include exactly one of source or sources
• use type: "event" for each source
• include aggregation
• only use supported filter operators and chart types
• if the user wants to group by or filter by event value, use !statsig_value rather than value

1. Execute the matching script.
2. Return the API response, or summarize the created dashboard ID or widget IDs if that is all the user needs.

Service Filters

When a widget needs to scope to a Statsig service, prefer filtering on:

{
  "property": {
    "key": "custom.service",
    "column": "user_object"
  },
  "operator": "EQ",
  "values": ["service-name"]
}

Do not default to bare service when building dashboard widgets. In Statsig event payloads this is commonly nested under user_object.custom, so custom.service is the safer default.

OTEL Aggregations

When a widget targets a metric that may come from OTEL, sample the metric first and inspect company_metadata.__metric_type.

• If __metric_type is present, treat the metric as OTEL-backed and prefer the matching OTEL aggregation family instead of the plain aggregation.
• If __metric_type is absent on a representative sample, treat it as a non-OTEL metric and use the normal aggregation types.

Preferred mapping:

• __metric_type = "gauge": use OTEL_GAUGE__AVG, OTEL_GAUGE__MAX, OTEL_GAUGE__MIN, or OTEL_GAUGE__SUM
• __metric_type = "counter": use OTEL_COUNTER__SUM, OTEL_COUNTER__AVG, OTEL_COUNTER__MIN, or OTEL_COUNTER__MAX
• __metric_type = "histogram": use the OTEL histogram aggregations documented in references/dashboard-api.md

Keep the aggregation intent the same when converting:

• average-style widgets should become OTEL_GAUGE__AVG instead of AVG
• max-style widgets should become OTEL_GAUGE__MAX instead of MAX

Guardrails

• Do not invent unsupported widget types such as tables.
• Do not include layout fields when creating a dashboard. The create endpoint does not accept widget positions.
• Do not include row or column. The API currently supports widget sizing via width and height, not explicit placement.
• Do not include id in the create request body.
• If the user gives you a dashboard from GET /console/v1/dashboards/{id}, reuse that create-compatible shape but remove id.
• Do not use the create endpoint when the user wants to mutate an existing dashboard.
• Prefer markdown formatting in text widgets so the content is readable and structured.

Examples

Minimal dashboard:

cat > /tmp/dashboard.json <<'JSON'
{
  "name": "Checkout Overview"
}
JSON

python3 scripts/create_dashboard.py \
  --body-file /tmp/dashboard.json

Read an existing dashboard into a reusable create payload:

python3 scripts/read_dashboard.py \
  --dashboard-id dash_123

Append widgets to an existing dashboard:

python3 scripts/add_dashboard_widgets.py \
  --dashboard-id dash_123 \
  --body-file /tmp/add-widgets.json

Replace all widgets on an existing dashboard:

python3 scripts/replace_dashboard_widgets.py \
  --dashboard-id dash_123 \
  --body-file /tmp/replace-widgets.json

Script Notes

• All scripts read STATSIG_CONSOLE_API_KEY from the environment.
• All scripts default STATSIG_API_VERSION to 20240601.
• Pass --api-version or set STATSIG_API_VERSION to override that default.
• All scripts default to https://api.statsig.com/console/v1/dashboards.
• All scripts send the Console API key in the statsig-api-key header.
• All scripts send the configured statsig-api-version header.
• create_dashboard.py, add_dashboard_widgets.py, and replace_dashboard_widgets.py accept:
• --body-file
• --body
• stdin
• read_dashboard.py, add_dashboard_widgets.py, and replace_dashboard_widgets.py require --dashboard-id.
• All scripts support --dry-run.
• All scripts support --raw.
• read_dashboard.py prints the raw GET response only when --raw is set. Otherwise it unwraps data and removes id.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: statsig-io
• Source: statsig-io/agent-skills
• License: ISC

Install and usage instructions live in the source repository linked above.