Jira Service Management (JSM) Assets Extension

This extension provides an MCP server for interacting with Jira Service Management (JSM) Assets.

Project Structure

The codebase is a Python package (jsm_asset_mcp/) with the following modules:

Module	Responsibility
`config.py`	`Settings` dataclass loaded from env vars, with cached/discovered Jira IDs
`client.py`	`AssetsClient` — thin httpx wrapper with auth and base URL
`cache.py`	Generic `TTLCache` — domain-agnostic
`schema.py`	`SchemaService` — schema introspection and summary builder
`llm.py`	Claude Agent SDK / Gemini structured-output natural-language → AQL translation
`tools.py`	All MCP tool definitions (thin orchestration)
`server.py`	`create_server()` factory — wires dependencies, returns FastMCP

Entrypoint: main.py calls create_server().run().

Query Assets: Use execute_aql to query objects based on Asset Query Language (AQL).
Search Assets: Use search_assets for natural-language object searches. The configured LLM decides the AQL query, result type, and result limit; count questions use the Assets total-count endpoint, and all-object requests paginate through all AQL result pages.
Retrieve Assets: Use get_object to fetch specific details about an asset by ID.
Update Assets: Use update_object to modify an existing asset's attributes.

Always construct an accurate AQL query when searching.
Ensure you have the object_type_id when updating an asset.
When you update an object, follow the attribute structure strictly (i.e., array of {"objectTypeAttributeId": "...", "objectAttributeValues": [{"value": "..."}]}).