Lentidão na consulta de informações MCP JIRA via copilot studio

Julia Angerami da Silva

Olá,

Identificamos que o MCP JIRA ITSM performa com maior lentidão (~8 minutos) na ferramenta Copilot Studio, quando comparado ao uso nas ferramentas Claude e GitHub Copilot (~1 minuto).
Alguém têm experienciado algo parecido? Conhecem alguma solução?

Tomislav Tobijas

Hmm @Julia Angerami da Silva 🤔

I'm not sure if this is still the case, but Jira MCP tools did fetch full issue/work item objects, including all metadata and custom fields. Even for simple queries…

Could be that Copilot processes thresholds differently or has client-side handling for large JSON payloads, and that could result in significant latency or "freezing" during the query process. 👀

But, I'm no expert in this and haven't actually used Copilot Studio to be able to confirm.

As for some potential solutions, you could try these (AI-generated btw):

Narrow JQL Queries: Modify your prompts to be highly specific. Instead of general searches, include date ranges (e.g., created >= "-7d") or filter by specific statuses and assignees to limit results to 10 issues or fewer

Avoid Count-Only Queries: Do not ask the AI to "count" issues. The MCP server will still fetch the full data for every matching issue just to count them. Use the native Jira interface for large-scale counts

Update Endpoints: Ensure you are using the most current endpoint: https://mcp.atlassian.com/v1/mcp/authv2. Legacy endpoints (like those ending in /sse) are being deprecated and can cause "invalid context" errors or performance bottlenecks

Verify Authentication: Ensure the connection uses OAuth 2.1 with Dynamic Client Registration (DCR). Manually filling in OAuth fields can sometimes lead to internal policy blocks that slow down tool execution

What I'd maybe recommend is reaching out to Atlassian Support if you're eligible to see if their teams have any additional insights 🎟️

Cheer,
Tobi