Marxel MCP Server 是官方的吗?
是的。服务器自称 Marxel,指向厂商官网 www.marxel.co,并运行在 Marxel 的域名上。
@Marxel
基本信息
Marxel MCP 服务器让 AI 助手按招聘人员确认的标准评审最多 5 份用户提供的简历,引用每份简历中的支撑证据,并给出用于核实的面试问题。选定的评审可以作为 Screening Job 保存到 Marxel 组织。它从不对候选人排名,也不做录用决定。
https://mcp.marxel.co/mcpclaude mcp add marxel --transport http https://mcp.marxel.co/mcpOpen the Marxel recruitment workspace with the labelled fictional sample. Returns the sample hiring brief and criteria, every sample candidate with a status, supporting detail, source and interview question per criterion, and the selected candidate IDs. Do not make or execute a hiring decision.
Compare candidates from the labelled fictional sample against its hiring brief. Returns the brief and its criteria, every sample candidate with a match, partial or missing status, supporting detail, source and verification interview question for each criterion, and the IDs selected for comparison. It does not compare user-supplied CVs; use create_review_workspace for those. This tool is read-only.
Build a candidate comparison and interview-question workspace from user-supplied CVs. Before calling: call confirm_criteria and use exactly the criteria it returns as confirmed, and ask the user to confirm the candidate batch; a prior confirmation does not cover a changed set. Each workspace supports at most 5 candidates and 8 criteria. With more than 5 CVs, ask the user to choose a batch of up to 5 and name the deferred candidates; never silently omit candidates or criteria. If confirming criteria in chat and there are more than 8, propose a reduced or consolidated set for the user to confirm. Distinguish explicit requirements from responsibilities and desirable experience: use Responsibility for role duties, and do not make them Must-have without the user's agreement or relabel them Nice-to-have. Quote exactly and verbatim from each candidate's sourceText; use match only when the quote supports the full criterion, and partial when it covers only part. Do not rank candidates, assign Screening Buckets, or make hire or reject decisions. To correct an assessment, call this tool again with the corrected status, all original source text, and the same criteria and candidate batch before claiming the workspace is updated; if that fails, say the visible workspace still shows the previous assessment. The latest successful workspace replaces earlier ones for discussion and export. Returns the brief, each candidate's status, quote, source and interview question per criterion, and any unreadable CVs. A quote not found verbatim fails the call. Stores nothing and changes no Marxel records; save_to_marxel imports a review.
Ask the recruiter to confirm the job-related criteria before any evidence is extracted. Propose up to 12 criteria derived from the hiring brief, with Must-have only for explicit requirements, Responsibility for role duties and Nice-to-have for desirable experience. Where the client supports forms, the recruiter reviews them in a form and keeps at most 8; otherwise confirm them in chat. Returns a state of confirmed with the recruiter's kept criteria and weights, not_confirmed when the form was closed, or confirm_in_chat with the proposed criteria when forms are unavailable. Use only the criteria this tool returns as confirmed, and do not create a workspace when it reports they were not confirmed.
List the Marxel organisations the connected account can use for a review save. Returns each organisation's ID and name, plus the account's active organisation ID when there is one; pass the chosen ID to save_to_marxel. This is read-only and does not create or change any records.
Save the selected review to the explicitly chosen Marxel organisation after the user confirms the import. This creates a Screening Job and selected Candidate records and starts screening; it does not send messages or make hiring decisions. Reuse the same UUID idempotencyKey on retries of the same confirmed review; generate a new key if the review or selection changes.
Check the result and current plan details for a previously confirmed Marxel review save. Returns links to the saved Screening Job, team settings and plan page, the number of candidates imported, and plan details when available. This is read-only; use the same organisation and idempotency key from the save request.
Marxel MCP Server 是 Marxel(一款内置 AI 简历筛选的招聘管理系统)提供的托管 MCP 端点,可让 AI 助手按已确认的招聘标准比较简历、引用支撑证据并准备面试问题。
该服务器(名为 marxel-shortlist-studio)围绕“人工参与”的评审流程设计:招聘人员先确认标准,助手从每份简历中摘录原文引用,结果可以作为 Screening Job 保存到招聘人员的 Marxel 工作区。它自身的说明要求助手绝不对候选人排名、分配筛选分组,也不做录用或拒绝决定。
端点为 https://mcp.marxel.co/mcp,传输方式为 Streamable HTTP。我们的探测无需登录即可连接并列出工具,因此添加后即可试用评审流程,包括一个带标注的虚构示例。保存评审时会使用已连接账号可用的 Marxel 组织,因此这一步需要 Marxel 账号。根据其定价页面,Marxel 提供 Free 计划(每月 200 份简历)以及付费计划。
claude mcp add --transport http marxel https://mcp.marxel.co/mcp
打开 Settings,进入 Connectors,选择 Add custom connector,名称填写 Marxel,URL 粘贴 https://mcp.marxel.co/mcp。
将服务器添加到 ~/.cursor/mcp.json(或项目内的 .cursor/mcp.json):
{
"mcpServers": {
"marxel": {
"url": "https://mcp.marxel.co/mcp"
}
}
}
添加到 .vscode/mcp.json:
{
"servers": {
"marxel": {
"type": "http",
"url": "https://mcp.marxel.co/mcp"
}
}
}
典型流程:粘贴职位说明和最多 5 份简历,让助手提出标准并由你确认,查看证据工作区,再决定是否保存。
探测返回了 7 个工具,可分为三组:
confirm_criteria 请招聘人员确认标准;create_review_workspace 仅使用已确认的标准,基于用户提供的简历构建候选人比较与面试问题工作区。open_shortlist 和 compare_candidates 只作用于带标注的虚构职位说明和候选人,适合预览输出格式。compare_candidates 为只读。list_marxel_organizations 列出已连接账号可用的组织;save_to_marxel 保存已确认的评审(创建 Screening Job 和所选候选人并开始筛选);get_marxel_save_status 返回已保存职位的链接、导入的候选人数量和计划信息。confirm_criteria 会返回 not_confirmed;请重新确认或在对话中确认。list_marxel_organizations 只返回已连接账号可用的组织。请确认你有 Marxel 账号并拥有该组织的访问权限。get_marxel_save_status 查看结果。是的。服务器自称 Marxel,指向厂商官网 www.marxel.co,并运行在 Marxel 的域名上。
我们的探测无需登录即可连接并列出工具。将评审保存到 Marxel 时会使用已连接账号的组织,因此这一步需要 Marxel 账号。
Marxel 定价页面列出了每月 200 份简历的 Free 计划,以及付费的 Standard、Plus 和 Business 计划。最新信息请查看定价页面。
不会。它的说明禁止对候选人排名、分配筛选分组或做录用与拒绝决定,只展示证据供人工评审。
每个评审工作区最多支持 5 位候选人和 8 条标准,更多候选人需分批处理。
支持。它是标准的 Streamable HTTP 端点,任何支持远程 MCP 服务器的客户端都可以通过 `https://mcp.marxel.co/mcp` 连接。
`open_shortlist` 和 `compare_candidates` 使用带标注的虚构职位说明和候选人,让你无需上传真实简历就能看到输出效果。
Marxel 的远程 MCP 服务器是一个托管的 Model Context Protocol 端点,地址是 https://mcp.marxel.co/mcp,AI 助手无需在本地安装或运行任何东西即可连接。
把端点 https://mcp.marxel.co/mcp 添加到任意支持 MCP 的客户端,比如 Claude Code、Cursor 或 VS Code。本页的接入代码块可以一步配置好每个客户端。
此收录尚未确认 Marxel 的认证要求。连接前请查阅提供方文档。
Marxel 暴露的是 streamable-http 端点,这是远程 MCP 服务器普遍使用、主流客户端都支持的传输方式。
AI Video MCP by AITuber is a hosted MCP server that lets AI agents create and publish video. You connect it once, then ask your agent for a video in plain language. It writes the script, picks a voice, generates the visu
Pharmaceutical drug & clinical trial intelligence
Ideate, create, and deliver with Adobe pro tools
Search and read public data from X/Twitter, Reddit, YouTube, Instagram, TikTok, Facebook, Threads, Bluesky, Trustpilot, Amazon, Google Maps, news, forums and the web. 90+ read-only tools plus ready-made research playbook
Convert web pages to clean Markdown and retrieval-ready chunks with browser rendering support.
评论