mcp-server-atlassian-jira
An MCP server for interacting with Atlassian Jira, enabling issue management, attachments, and other operations via natural language.
README
mcp-server-atlassian-jira (UFC fork)
Private UFC fork of
aashari/mcp-server-atlassian-jira
(forked at upstream commit aab8b7f2, one commit past v3.3.0), carrying:
- Binary response corruption fix - upstream's
fetchAtlassian()(src/utils/transport.util.ts) read every response body viaresponse.text(); for binary endpoints (attachment content, thumbnails, exports) that is a lossy, irreversible UTF-8 decode. Fixed: non-JSON/text/XML bodies are read viaarrayBuffer()and returned as{ __binary: true, contentType, byteLength, base64 }. Covered by regression tests insrc/utils/transport.util.test.ts. - Ported upstream community PR #173 (author
cedral) -jira_attach(multipart upload) andjira_get_attachment(byte-exact download to a local file). DOTENV_CONFIG_PATHwith~-expansion insrc/utils/config.util.ts@loadFromEnvFile, plusUSE_DOTENVtoggle - credentials live in a per-user file outside any repo/package tree (team convention:~/.claude/jira.env).- Unconditional
main()insrc/index.ts(the upstreamrequire.main === moduleguard broke startup under embedded Node runtimes). node:httpstransport fallback (safeFetchinsrc/utils/transport.util.ts) - in some MCP host processes the globalfetch(undici) fails withTypeError: fetch faileddespite a live network (observed 2026-07-14 in a Claude-Code-session-spawned connector instance, while a Desktop-spawned instance on the same machine was fine). On that failure the request is retried through the classicnode:httpsstack (multipart encoded vianew Response(FormData), redirects followed withAuthorizationdropped cross-host).FORCE_HTTPS_FALLBACK=trueforces the fallback path (diagnostics / emergency lever). Verified live: JSON GET, JQL search, and a byte-exact 2.1 MB binary download through the media-CDN redirect.- Unresolved-template env guard + creds-path fallback - some hosts pass
the MCPB manifest env without substituting
${user_config.*}; a literal "${...}" value is non-empty and would shadow the .env file and leak into the URL ("ENOTFOUND ${user_config...}.atlassian.net"). Such values are now treated as unset, and when noDOTENV_CONFIG_PATHsurvives, the .env lookup falls back to<cwd>/.env, then~/.claude/jira.env.
Full background and verification evidence:
tools/jira-mcp-plugin/README.md in the 1C_Workspace repo.
How this repo is consumed
This repo exists as a standalone git package so that npx can install it
directly (npx cannot install from a monorepo subdirectory). The
jira@ufc-1c Claude Code plugin (marketplace arcankostenko/1C_Workspace)
declares:
"command": "npx",
"args": ["-y", "github:asky74/mcp-server-atlassian-jira#v3.3.0-ufc.3"]
On first start npm clones this repo (using your system git - your existing
GitHub credentials cover the private access), installs devDependencies, and
the prepare script compiles TypeScript into dist/. Subsequent starts run
from the npx cache. No manual npm install step.
Releases are tags (v<upstream>-ufc.<n>, e.g. v3.3.0-ufc.1). To ship an
update: commit, tag, push the tag, then bump the tag in the plugin's
.mcp.json.
Local working-clone convention: 1C_Workspace/mcp-servers/mcp-server-atlassian-jira/
- an inner repo like
BAS_KUP_local/(own.git/origin, not tracked by the monorepo's whitelist.gitignore).
Credentials: ~/.claude/jira.env with ATLASSIAN_SITE_NAME,
ATLASSIAN_USER_EMAIL, ATLASSIAN_API_TOKEN (see .env.example). Never
commit credentials here.
Development
npm install # also builds (prepare -> tsc)
npm test # 6 suites / 61 tests
manifest.json + .mcpbignore here build the optional Claude Desktop MCPB
bundle (npx @anthropic-ai/mcpb pack .) - a personal, per-machine option.
The 2026-07-13 "built-in-node fetch blocker" is not reproducible on
current Desktop (>=1.20186, Electron 42.5.1 / node 24.17): a probe script
run inside the real MCP runtime on 2026-07-14 got HTTP 200 from both undici
fetch and classic https.get, and the packed bundle works with the stock
command: "node" - no node.exe workaround needed. .mcpbignore caveat:
patterns are gitignore-style, keep them anchored (/src/, not src/) -
an unanchored src/ strips node_modules/*/src and breaks the bundle.
Licensing
Upstream declares "license": "ISC" in package.json but ships no LICENSE
file. This fork keeps the declared license and upstream attribution
(original work: Andi Ashari; attachment tools: cedral, PR #173).
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.