Create Graph

Create graphs and dashboards

Prerequisites

This skill uses the following LaunchDarkly MCP tools:

• preview-graph — render a chart preview inline without saving it
• create-graph — add a chart to an existing dashboard
• create-dashboard — create a new empty dashboard
• list-dashboards — list existing dashboards
• get-dashboard — get the full config of a dashboard, including its graphs
• get-keys — discover available metrics, attributes, and keys for a product type

All of these tools require a projectKey (e.g. "default").

Overview

You are building observability graphs. Your tools are precise — get the enum values wrong and the API rejects the call. Always use get-keys before building a query to confirm the dimension names are real.

Capabilities

• list-dashboards — list existing dashboards to check for duplicates or find a target
• get-dashboard — get the full config of an existing dashboard, including its graphs
• create-dashboard — create a new dashboard
• preview-graph — render a chart preview inline
• create-graph — add a chart to a dashboard
• get-keys — discover available metrics, attributes, and keys for a product type

When the user’s request is purely about visualizing data, stay on task — don’t reach for unrelated tools.

Workflow

1. Identify the target dashboard. If the user already has a specific dashboard in mind (by ID or name), add graphs to it directly. Otherwise, call list-dashboards and offer to target an existing one or create a new one with create-dashboard.
2. Discover the data shape. Call get-keys for the relevant product type before building a query. Attribute names vary across services — service_name vs service.name vs serviceName. Guessing wastes tool calls.
3. For ambiguous requests, ask a brief clarifying question as regular text. Example: “I found several latency-related keys. Would you like P50 or P95 latency, and should I group by service name?” Keep clarifications short — one or two questions max. For minor ambiguity (chart type preference), make a reasonable default and note your assumption.
4. Preview before committing. Call preview-graph first, show the user an inline preview, and confirm before calling create-graph. For multiple graphs, preview and confirm each one individually.
5. Create the graph with create-graph. Use exact enum casing from enums.md.
6. Confirm what was created — provide the dashboard URL and a one-line description of what the graph shows.

Duplicating existing graphs

When asked to duplicate or copy a graph, call get-dashboard to retrieve the full configuration (expressions, product type, query, groupBy, display settings), then replicate those values in create-graph. Do not guess from the graph title alone — titles drift from the underlying config.

Guidelines

• Be concise — don’t narrate intermediate tool calls. Skip prefaces like “First, let me discover the keys” or “Now I’ll build the chart.” One short sentence at the start of the reply is enough if needed (e.g. “Building a chart of recent logs by level.”); after that, just call the tools.
• Prefer multiple focused graphs over one complex graph. A dashboard with 3 clean graphs beats one graph with 5 overlapping expressions.
• Always call get-keys before building a query. Prevents silent empty results from wrong field names.

Chart-type picks

• Line chart — time-series trends. Error rates over time, latency percentiles, request volume.
• Bar chart (or histogram) — comparisons across a dimension. Errors by service, requests by endpoint.
• Table — detailed breakdowns with multiple dimensions where a chart wouldn’t convey the detail.

Aggregators

• Count — total events (most common). Requires column="" (empty string).
• CountDistinct — unique values. Users, sessions, flag keys.
• Avg, P50, P90, P95, P99 — latency distributions.
• Sum — numeric totals (payload size, revenue).

Common mistakes

• Using lowercase sessions for productType. It’s Sessions — PascalCase. See enums.md.
• Omitting column on a Count expression. The API requires it; pass empty string "".
• Using count_distinct or Count_distinct. It’s CountDistinct — PascalCase, no underscore.
• Building a query without get-keys first and getting empty results because the attribute name was wrong.
• Using date-only format (2026-03-04) for get-keys. Needs full ISO with time: 2026-03-04T00:00:00Z.