--- title: Jira’s depth, on demand: designing an MCP server around... description: Atlassian redesigned its Jira MCP server around jobs to be done instead of mapping tools directly to REST endpoints. The v1 version exposed 15 Jira tools tied... canonical: https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Jira’s depth, on demand: designing an MCP server around jobs, not endpoints | daily.dev og:description: Atlassian redesigned its Jira MCP server around jobs to be done instead of mapping tools directly to REST endpoints. The v1 version exposed 15 Jira tools tied... og:url: https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe og:image: https://api.daily.dev/og/posts/fUVS6Pgoe.png og:image:alt: Jira’s depth, on demand: designing an MCP server around jobs, not endpoints og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Jira’s depth, on demand: designing an MCP server around jobs, not endpoints **[Atlassian](https://daily.dev/sources/atlassian)** · 15 min read · 0 upvotes · 0 comments ## Summary Atlassian redesigned its Jira MCP server around jobs to be done instead of mapping tools directly to REST endpoints. The v1 version exposed 15 Jira tools tied closely to API calls; v2 ships 52 tools that together call over 100 APIs, grouped by intent like read_jira and write_jira. v2 loads 6 tools upfront, with the rest deferred, and saves up to 25% in tokens overall. In internal testing, failed tool calls per run dropped from 0.20 to 0.09, and v2 passed 57 of 62 shared test cases versus 54 for v1. The Atlassian MCP server now has nearly 2M monthly active users, up 8x since March 1. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Should an MCP server expose one tool per REST API endpoint or group tools around user tasks? Atlassian found that grouping tools around jobs to be done, rather than one tool per endpoint, improved reliability and reduced cost. Mapping Jira's roughly 725 REST operations directly would have created tool sprawl, more wrong tool picks by agents, and schema bloat, since most users only exercise a small slice of the API anyway. Their curated, job-based v2 design has 52 tools covering over 100 APIs. _Weighing tool-per-endpoint versus task-based MCP design is easier with real comparisons surfaced on daily.dev._ ### How much do agent tool failure rates improve when an MCP server reduces the number of loaded tool schemas? Atlassian's Jira MCP v2 cut failed tool calls from 0.20 to 0.09 per run compared to v1. v2 loads only 6 always-available tools instead of v1's 32 upfront tool definitions (about 15k tokens), deferring the remaining 46 tools behind a discover mechanism, and this also cut token usage by up to 25% for the same work. _Developers tuning agent reliability can follow MCP server design tradeoffs like these on daily.dev._ ### How can an MCP server keep token usage low while still exposing a large API surface to an AI agent? Atlassian's approach loads only 6 core tools upfront and defers the other 46 behind discover, executeRead, executeWrite, and executeDestructive calls, so risk tier is explicit per call. In a separate architecture comparison, loading the full tool catalog upfront used roughly 2.2 times the cumulative input tokens versus this curated, deferred approach. _Anyone architecting MCP servers for large APIs can track token-saving patterns like this on daily.dev._ ## Similar posts on daily.dev - [Atlassian MCP brings more of your work within reach of any AI agent](https://daily.dev/posts/atlassian-mcp-brings-more-of-your-work-within-reach-of-any-ai-agent-uusqnay4r) · Atlassian · 1 upvotes · 0 comments - [10 Interesting MCP Statistics](https://daily.dev/posts/10-interesting-mcp-statistics-wxppqen4j) · Nordic APIs · 0 upvotes · 0 comments - [What 5M\+ daily MCP tool calls taught us about the future of AI at work](https://daily.dev/posts/what-5m-daily-mcp-tool-calls-taught-us-about-the-future-of-ai-at-work-4vplnrnuq) · Atlassian · 1 upvotes · 0 comments - [Why More MCP Tools Can Make an AI Agent Less Useful](https://daily.dev/posts/why-more-mcp-tools-can-make-an-ai-agent-less-useful-piqjmps6m) · Dev Genius · 3 upvotes · 1 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#architecture](https://daily.dev/tags/architecture), [#mcp](https://daily.dev/tags/mcp), [#atlassian](https://daily.dev/tags/atlassian), [#jira](https://daily.dev/tags/jira) [View this post on daily.dev](https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Jira’s depth, on demand: designing an MCP server around jobs, not endpoints","url":"https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe"},"datePublished":"2026-10-28T22:02:57.723Z","dateModified":"2026-10-28T23:08:19.447Z","description":"Atlassian redesigned its Jira MCP server around jobs to be done instead of mapping tools directly to REST endpoints. The v1 version exposed 15 Jira tools tied...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d92bad28cc3fcb901f80fdff712229ab?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d92bad28cc3fcb901f80fdff712229ab?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Atlassian","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Atlassian","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/44750d8e66de4411bb477159b321bb43","url":"https://daily.dev/sources/atlassian"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,architecture,mcp,atlassian,jira","timeRequired":"PT15M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Atlassian","item":"https://daily.dev/sources/atlassian"},{"@type":"ListItem","position":3,"name":"Jira’s depth, on demand: designing an MCP server around jobs, not endpoints"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/jira-s-depth-on-demand-designing-an-mcp-server-around-jobs-not-endpoints-fuvs6pgoe#faq","mainEntity":[{"@type":"Question","name":"Should an MCP server expose one tool per REST API endpoint or group tools around user tasks?","acceptedAnswer":{"@type":"Answer","text":"Atlassian found that grouping tools around jobs to be done, rather than one tool per endpoint, improved reliability and reduced cost. Mapping Jira's roughly 725 REST operations directly would have created tool sprawl, more wrong tool picks by agents, and schema bloat, since most users only exercise a small slice of the API anyway. Their curated, job-based v2 design has 52 tools covering over 100 APIs. Weighing tool-per-endpoint versus task-based MCP design is easier with real comparisons surfaced on daily.dev."}},{"@type":"Question","name":"How much do agent tool failure rates improve when an MCP server reduces the number of loaded tool schemas?","acceptedAnswer":{"@type":"Answer","text":"Atlassian's Jira MCP v2 cut failed tool calls from 0.20 to 0.09 per run compared to v1. v2 loads only 6 always-available tools instead of v1's 32 upfront tool definitions (about 15k tokens), deferring the remaining 46 tools behind a discover mechanism, and this also cut token usage by up to 25% for the same work. Developers tuning agent reliability can follow MCP server design tradeoffs like these on daily.dev."}},{"@type":"Question","name":"How can an MCP server keep token usage low while still exposing a large API surface to an AI agent?","acceptedAnswer":{"@type":"Answer","text":"Atlassian's approach loads only 6 core tools upfront and defers the other 46 behind discover, executeRead, executeWrite, and executeDestructive calls, so risk tier is explicit per call. In a separate architecture comparison, loading the full tool catalog upfront used roughly 2.2 times the cumulative input tokens versus this curated, deferred approach. Anyone architecting MCP servers for large APIs can track token-saving patterns like this on daily.dev."}}]} ```