MSPortfolio — MCP-Native Engineering Portfolio

io.github.ManSio由 GitHub 账号 ManSio 发布v1.0.4更新于 Oct 24, 2026

MCP server for Mikhail (ManSio)'s engineering portfolio: projects, principles, lab, evidence.

发布方自己的服务：PortfolioFit

已验证Streamable HTTP可网页运行知识与记忆数据与分析

服务详情

标识符: io.github.ManSio/msp-portfolio
传输: Streamable HTTP
运行时: Web executable
信任: Verified
发布于: Sep 16, 2026

安全提示

MCP 服务会接收对话中的工具调用请求，并可能执行外部操作。启用前请先了解该服务能做什么，并只授予它所需的权限。

注册中心已验证此条目所属的 GitHub 账号 ManSio。这并不表明该账号就是产品方。

SourceWeft 的检查只涵盖条目是否完整、端点是否响应；不审查服务器的代码，也不审查其工具的行为。

在 README 中添加徽章

向读者展示该 MCP 服务已收录于 SourceWeft，徽章链接到本页。

[![Listed on SourceWeft](https://sourceweft.com/badge/mcp/io.github.ManSio%2Fmsp-portfolio.svg)](https://sourceweft.com/mcp/io.github.ManSio%2Fmsp-portfolio?utm_source=readme&utm_medium=badge)

概览

AI 生成的概览

让助手查询并核对这位工程师的作品集资料，包括项目、原则、实验与已知问题。

可以做什么: 提供一组只读工具，用于查询这位工程师的个人作品集：个人简介、项目与决策记录、工程原则、时间线、文章、提交历史、反模式、实验、日志与已知问题看板。另有 analyze_stack 比对技能需求、simulate_architecture 模拟架构在负载等场景下的表现，以及 verify_claim、verify_repo、verify_article、verify_package 等工具，用外部来源核对相关说法。
适用场景: 当您想用可核实的资料回答关于这位工程师的问题时使用，例如面试准备、项目审查，或核对关于其项目、文章与软件包的说法。
运行要求: 远程 HTTP 端点，无需安装软件包、环境变量或身份验证。部分核对工具会连接 GitHub、Dev.to 与 npm registry 等外部 API，需要网络访问。

安装前请注意

多数工具为只读，但 verify_repo、verify_article 与 verify_package 会向 GitHub、Dev.to 及 npm registry 发出查询，可能暴露您所查询的内容。数据为部署时的快照，可能不是最新。

安装

在 SourceWeft 中

打开控制台中的 MSPortfolio — MCP-Native Engineering Portfolio，将其添加到工作区。
为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.ManSio%2Fmsp-portfolio/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "msportfolio-native-engineering-portfolio": {
      "type": "http",
      "url": "https://msp-portfolio.mansio-dev.workers.dev/mcp"
    }
  }
}

工具

get_profile

Get the owner's professional profile summary, plus nextSteps — concrete ways to continue (contact channels, GitHub, MCP connect). Use it as the first tool in an interview.

无需输入参数。

get_projects

Get portfolio projects with stack, highlights, and decision logs. Optional filter by stack tag.

输入参数filter

search_projects

Free-text search across portfolio projects — name, tagline, description, stack and decision-log rationale. Returns matched projects with a score and the fields that matched. Use when a fixed stack-tag filter is too narrow (e.g. 'RAG', 'latency', 'Zed'). Read-only, closed world.

输入参数query必填

get_engineering_principles

Get engineering principles with real examples and A/B-style counterfactuals.

无需输入参数。

get_timeline

Get the engineering decision timeline.

无需输入参数。

get_articles

Get recent Dev.to articles with reading time, tags and links.

无需输入参数。

get_commit_history

Get recent commit history across the owner's public repos. Reads the deployed snapshot and reports its `fetchedAt` and `ageMinutes`, so 'how current is this?' is answered by the payload instead of by trusting this description. Use it to answer 'what has he been building lately' or 'show the hardest bugs he has fixed'.

无需输入参数。

get_antipatterns

Get the owner's antipattern museum — real engineering mistakes with why they were bad, how they were fixed, and the lesson. Read-only, closed world.

无需输入参数。

analyze_stack

Compare the owner's stack against a job's required skills. Returns per-skill match with evidence and coverage.

输入参数required_skills必填

simulate_architecture

Simulate how a project's architecture behaves under a scenario (load spike, node loss, cache cold, LLM saturation). Returns latency percentiles per load and bottleneck analysis.

输入参数scenario必填project_id必填

get_experiments

Get the owner's engineering experiments — hypothesis, command, raw result and verdict (confirmed/refuted/partial) for each. Use it to answer 'what did you measure' or 'show me an experiment you ran'. Includes negative results (approaches that failed). Read-only, closed world.

无需输入参数。

get_diary

Get the owner's engineering diary — incidents, root causes, fixes and guards, each tagged with a pattern (NEW vs recurring). Use it to answer 'what broke and how did you fix it' or 'show your hardest debugging session'. Read-only, closed world.

无需输入参数。

get_known_issues

Get the owner's known-issues board — open debt with status, temperature (stable/watching) and deadlines. Use it to answer 'what's still broken' or 'what are you working on'. Read-only, closed world.

无需输入参数。

get_issue_detail

Get the full detail of a single known issue by its ID (e.g. 'KI-109'). Drills into one open problem: status, temperature, owner, linked source and the raw problem statement. Returns a clear not-found with the available IDs if the ID does not match. Read-only, closed world.

输入参数id必填

verify_claim

Ground a claim about the owner against the portfolio's data (profile, projects, principles, timeline, antipatterns, experiments, diary, known issues). Deterministic: returns the evidence records that support the claim (with source paths) and a supported verdict. Use it before asserting a fact about the owner, or to check what an answer was based on.

输入参数claim必填

verify_repo

Verify a GitHub repository against the primary source: fetches the actual repo metadata (exists, language, description, topics, stars, last push) from the GitHub API and cross-checks it with the portfolio's project record when the repo is one of the owner's projects (language/stack agreement). With readme:true it also returns the actual README text, so claims about what the project does can be checked against the repository's own words. Use it to ground claims like 'the repo is Python' or 'he maintains mscodebase-intelligence' with live data instead of trusting the claim. Read-only, open world (network fetch).

输入参数repo必填readme

verify_article

Verify an article against the primary source (Dev.to): does the owner have a published article matching the query? Fetches the live Dev.to API for the owner's articles and returns the real title/date/reactions/url when found — or an honest 'not found'. Use it to ground claims like 'he wrote about agent memory' in the platform's data instead of the portfolio's own words. Read-only, open world (network fetch).

输入参数query必填

verify_package

Verify an npm package against the primary source (registry.npmjs.org): does it exist, latest version, publish date, description, license, maintainers — and is the owner among them? Use it to ground claims like 'he published an npm package' in the registry's data instead of the portfolio's. Honest 'not found' when the package does not exist. Read-only, open world (network fetch).

输入参数package必填