Jira Product Discovery MCP
Exposes generic Jira REST tools and native Jira Product Discovery operations for Jira Cloud, including insights and attachments.
README
Jira Product Discovery MCP
An MCP server for Jira Cloud and Jira Product Discovery. It exposes the existing generic Jira REST tools together with typed native JPD Insight and attachment operations.
Package: @enterpret/jira-product-discovery-mcp
Tool surface
The server exposes exactly eight tools:
| Tool | Purpose |
|---|---|
jira_get |
Read from a Jira REST endpoint. |
jira_post |
Create through a Jira REST endpoint. |
jira_put |
Replace through a Jira REST endpoint. |
jira_patch |
Partially update through a Jira REST endpoint. |
jira_delete |
Delete through a Jira REST endpoint. |
jira_add_attachment |
Upload one UTF-8 or base64 attachment to an issue or JPD idea. |
jira_list_jpd_insights |
List native JPD Insights for an idea key. |
jira_create_jpd_insight |
Create a native JPD Insight with linked evidence in its description. |
The generic tools preserve compatibility with Jira REST API v3. Callers provide only the REST path, query parameters, and request body; authentication remains server-owned.
Authentication
OAuth bearer mode
Enterpret Agent uses Nango to refresh the Jira 3LO connection and injects the short-lived access token into the MCP subprocess:
ATLASSIAN_OAUTH_BEARER=<access-token> \
npx -y @enterpret/jira-product-discovery-mcp@0.2.0
Native JPD Insight tools require this mode. They do not accept tokens, Atlassian IDs, GraphQL, or OAuth client IDs as tool arguments.
Set ATLASSIAN_CLOUD_ID when the bearer can access more than one Atlassian
site. The server verifies the configured cloud ID against Atlassian's accessible
resources and otherwise fails closed on ambiguous multi-site connections.
Jira API token mode
The generic Jira REST tools and attachment tool also support Jira API-token authentication:
ATLASSIAN_SITE_NAME=your-instance
ATLASSIAN_USER_EMAIL=you@example.com
ATLASSIAN_API_TOKEN=<api-token>
This mode does not support native JPD Insights.
MCP configuration
The default transport is stdio:
{
"mcpServers": {
"jira-product-discovery": {
"command": "npx",
"args": ["-y", "@enterpret/jira-product-discovery-mcp@0.2.0"],
"env": {
"ATLASSIAN_OAUTH_BEARER": "<access-token>"
}
}
}
}
Supply secrets through the subprocess environment, never through MCP tool arguments or command-line flags.
Native JPD Insights
jira_list_jpd_insights accepts a human-friendly idea key such as MDP-2.
The server resolves the selected Atlassian cloud, project ID, and issue ID, then
constructs the fixed Polaris ARIs internally.
jira_create_jpd_insight accepts:
ideaKeydescriptionquotesourceUrl(HTTPS)sourceTitle
It creates a native JPD Insight with an Atlassian Document Format description
containing the summary, quote, and linked source. It does not create a structured
source snippet and does not require ATLASSIAN_OAUTH_CLIENT_ID.
Creation is intentionally marked non-idempotent. Retrying a create after an ambiguous provider response can create a duplicate Insight.
Native Insights currently use Atlassian's experimental Polaris GraphQL surface. The server exposes only fixed, typed list and create operations; arbitrary GraphQL and raw Polaris input are not available.
See JPD snippet representations for the difference between description-only Insights and structured source snippets.
Attachments
jira_add_attachment accepts an issue or idea key, filename, content, optional
encoding (utf8 or base64), and optional MIME type. The server builds Jira's
multipart request and limits decoded files to 10 MiB. Repeated calls create
separate Jira attachments.
Data handling
Generic Jira REST calls retain the inherited raw-response diagnostics under
/tmp/mcp/jira-product-discovery-mcp/. These files can contain Jira request and
response payloads and should be handled accordingly. Attachment uploads are
marked sensitive, so attachment bytes and returned content URLs are excluded
from debug and raw-response output.
Large generic Jira responses are truncated for the MCP caller and include the
corresponding raw-response path. Jira pagination, field selection, and the
generic tools' jq filter can be used to request a smaller result.
Development
Requires Node.js 18 or newer.
npm ci
npm run format:check
npm run lint
npm run build
npm test -- --runInBand
npm pack --dry-run
Tests use mocked provider responses for JPD and attachment operations. Do not use production Atlassian credentials for local test runs.
Releasing
Publishing is manual and never runs on a push to main. The release workflow
validates the candidate, packs once, smoke-tests that exact tarball through
npx, and publishes the same artifact with npm Trusted Publishing provenance.
See the release runbook for the one-time npm bootstrap, Trusted Publishing configuration, release gates, and post-publication verification.
License and ancestry
ISC licensed. This repository retains the generic Jira foundation and history from Andi Ashari's Jira MCP project and adds Enterpret's Jira Product Discovery integration. See LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。