Zenhub
Work with Zenhub boards from an AI agent through the GraphQL API: workspaces, pipelines and the issues in them, issue details, sprints and their issues, epics and their child issues, estimates; move issues, set estimates, plan sprints, create issues and comment. 17 typed tools plus a generic query tool, personal API key.
7 天免费 · 无需信用卡 · EU 托管
问 Zenhub 任何事
通过 AnythingMCP 对 Zenhub 试一个真实提示 — Claude 调用工具并返回答案。在你的消息上按 Enter 即可在 Cloud 安装。
在 Cloud 上安装 ZenhubClaude 是 AI，可能会出错。请核对回答。
2 分钟设置
- 1在 cloud.anythingmcp.com 开始免费试用。
- 2在连接器商店点击 Zenhub。提示时粘贴凭据。
- 3生成 MCP API 密钥，将 AI 客户端指向生成的 MCP URL。
Zenhub 指南
面向主流 AI 客户端的逐步设置。
配置 Zenhub
直接来自连接器定义：在哪里获取凭据、需要哪些权限，以及该适配器调用的是哪个 API。 由连接器作者以英文提供。
This connector uses the Zenhub GraphQL API (https://api.zenhub.com/public/graphql). Schema and examples: https://developers.zenhub.com/ (the Explorer shows the full schema).
Setup:
- In Zenhub open the dashboard's API section (https://app.zenhub.com/settings/tokens) and create a Personal API Key for the GraphQL API. A user can have up to 5 keys.
- Copy the key (it is shown once) into
ZENHUB_API_KEY.
Authentication: Authorization: Bearer <key>. A 401 means the key is wrong or revoked. The key acts as its user: workspaces and repositories that user cannot see are missing from every result.
Where to start: zenhub_get_viewer checks the key and lists the organizations. Find a workspace with zenhub_search_workspaces (by name) or zenhub_list_organizations. Then zenhub_list_pipelines for the board columns and zenhub_list_pipeline_issues for the issues in one column. Sprints: zenhub_list_sprints then zenhub_list_sprint_issues. Epics: zenhub_list_epics then zenhub_list_epic_issues.
Ids: Zenhub uses its own opaque ids (for example Z2lkOi8vcmFwdG9yL0lzc3VlLzEwNzQyOQ) for issues, pipelines, sprints, epics and repositories; workspace ids are 24-character hex strings. An issue number plus a repository identifies an issue in zenhub_get_issue; the mutations need the Zenhub issue id.
Limits: each API key gets 90 seconds of processing time per minute (token bucket, headers X-RateLimit-Limit, X-RateLimit-Used, X-RateLimit-Remaining) and at most 30 concurrent requests. A single query may not exceed a complexity of 200 points: every field counts 1 and a connection multiplies its children by first. That is why the list tools default to 10 to 30 items and request few fields. Avoid polling; Zenhub offers webhooks for change notifications.
Pagination: Relay cursors. Pass pageInfo.endCursor back as after while pageInfo.hasNextPage is true. first may not exceed 100, and the complexity limit usually bites first.
Errors: a GraphQL answer can be HTTP 200 with an errors list; the tool then reports the message. "Field doesn't exist" means the query does not match the schema.
Beyond the typed tools: every GraphQL connector also gets zenhub_graphql_query (read-only, any query document), zenhub_graphql_mutation, and a schema tool, for operations not covered here (releases, roadmaps, dependencies, priorities, labels).
Writes: zenhub_move_issue, zenhub_set_estimate, zenhub_add_issues_to_sprints, zenhub_create_issue and zenhub_add_comment change the board or post content visible to the team.