Collapsing each level to one tool with an operation argument keeps the tool list short, which matters when the interesting work is AQL anyway. The AQL syntax guide shipped as a resource is the detail that makes it usable — the model can read the query language before writing a query rather than guessing and getting a validation error back.
A server for Jira Insights, the asset and configuration database inside Jira Service Management. Three tools cover the three levels of the model — schema, object type, object — each taking an operation rather than splitting create, read, update and delete into separate tools.
- manage_jira_insight_schema — list, read, create, update and delete object schemas
- manage_jira_insight_object_type — the same operations against the object types inside a schema
- manage_jira_insight_object — objects themselves, plus a query operation that takes AQL
- jira-insights://instance/summary — top-level statistics about the instance
- jira-insights://aql-syntax — an AQL reference with examples, readable by the model before it writes a query
- jira-insights://schemas/all and per-schema resources for the full definition, overview and statistics
- jira-insights://object-types/{objectTypeId}/overview for one object type's attributes and counts
JIRA_HOST, JIRA_EMAIL and JIRA_API_TOKEN, with a Jira Insights instance the token can reach and permissions matching what you intend to do. Node.js 20 or later, or the published container image, which is the easier path for most people.
One command plus a key — docker run --rm -i -e JIRA_API_TOKEN -e JIRA_EMAIL -e JIRA_HOST ghcr.io/aaronsb/jira-insights-mcp:latest, then supply credentials