Skip to main content
Glama

DemandScope

公共 API 需求信号扫描器,用于验证数字产品和开发工具 创意,外加一个无依赖的 MCP(Model Context Protocol)stdio 服务器, 可将这些信号作为工具暴露给任何兼容 MCP 的代理调用。

透明度说明: 此项目由 在 Autonomous Income Research Lab 中运行的自主 AI 代理(Hermes,由 Nous Research 开发) 构建、测试和维护。人类所有者负责审查并授权外部操作。

为什么

由 Autonomous Income Research Lab(2026-08-17)构建,作为其第一个产品 原型:对小众开发工具的需求验证本身就是独立开发者反复出现的需求, 而 MCP 集成是一个快速发展的生态系统 (证据:research/2026-08-17-opportunity-scan.md)。

Related MCP server: AI Knowledge Center MCP

信号(全部免费、无需认证、对 ToS 友好的 API)

工具

来源

代理指标

github_repo_signal

api.github.com/search

开发者兴趣 / 竞争

github_trend

api.github.com (created:)

供应/兴趣加速(增量)

hn_signal

hn.algolia.com (Hacker News)

技术受众关注度

hn_trend

hn.algolia.com

关注度加速(增量)

npm_downloads

api.npmjs.org

可比产品需求

pypi_downloads

pypistats.org

可比产品需求

绝对下载量包含 CI/镜像流量——请将其用于方向性判断,而非作为 精确市场规模。趋势工具返回 growth_ratio(当前窗口 ÷ 前一个 窗口);>1 表示加速。

可靠性(v0.2)

  • 缓存: 基于文件的 TTL 缓存(DEMANDSCOPE_CACHE_DIR,默认 .cache/demandscope)。TTL:GitHub/HN 1 小时,npm/PyPI 24 小时。重复调用 免费且即时。

  • 退避: 在 HTTP 429/5xx 时进行指数重试(最多 4 次),并尊重 Retry-After。pypistats 被视为尽力而为。

用法

CLI 风格(批量扫描):

python3 ../demand-scanner/demand_scanner.py candidates.json out/

作为 MCP 服务器(通过 stdio 使用换行分隔的 JSON-RPC 2.0):

python3 mcp_server.py

MCP 客户端配置示例:

{"mcpServers": {"demandscope": {"command": "python3", "args": ["/path/to/mcp_server.py"]}}}

测试

python3 test_mcp_server.py

10 个端到端协议测试(握手、通知、ping、tools/list、 tools/call,包括趋势 + 缓存命中断言、错误路径)。 npm_downloads 和 hn_trend 调用测试需要网络。

状态 / 路线图

  • v0.2(2026-08-17): 趋势增量工具(github_trend、hn_trend)、TTL 缓存 + 速率限制退避、pyproject.toml 打包、MIT 许可证。 10/10 测试通过。

  • v0.1(2026-08-17):协议子集 initialize、ping、tools/list、 tools/call;4 个信号工具。

  • 下一步:resources 支持、托管 HTTP 传输、第二项资产 (mcp-registry-lint)在实验室流水线中。

Available Tools

6 tools
github_repo_signalA

GitHub repo count and top-star sums for a search query (developer interest / competition proxy).

Inputs: query (required).

github_trendC

GitHub repos created in the last N days vs the prior window; growth_ratio > 1 means accelerating supply/interest.

Inputs: query (required), window_days.

hn_signalA

Hacker News story count for a query over the last N days (tech-audience attention proxy).

Inputs: query (required), days.

hn_trendB

Hacker News story count for the last N days vs the prior window; growth_ratio > 1 means accelerating attention.

Inputs: query (required), window_days.

npm_downloadsB

Last-month downloads for an npm package (comparable-product demand).

Inputs: package (required).

pypi_downloadsA

Last-month downloads for a PyPI package (comparable-product demand).

Inputs: package (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.5/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each tool targets a distinct source (GitHub, Hacker News, npm, PyPI) and a distinct metric (absolute signal vs trend vs downloads). Even the GitHub and HN pairs are clearly separated by the signal/trend distinction, leaving no meaningful overlap.

    Naming Consistency4/5

    Most tools follow a source_metric pattern: github_repo_signal, github_trend, hn_signal, hn_trend, npm_downloads, pypi_downloads. The only minor inconsistency is that npm and PyPI use 'downloads' while GitHub and HN use 'signal'/'trend', but the pattern remains predictable and readable.

    Tool Count5/5

    Six tools is well-scoped for a demand-signal aggregator covering multiple external data sources. Each tool earns its place by representing a distinct signal-source combination, and the count feels neither too thin nor bloated.

    Completeness4/5

    The toolset covers both absolute and trend signals for GitHub and Hacker News, plus package downloads for npm and PyPI. A minor gap is the lack of trend/growth comparison for npm and PyPI downloads, but the core demand-signal surface is otherwise well covered.

    Maintenance

    ActivityStale
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      A local MCP server that gives AI agents access to developer tooling — GitHub (read-only), documentation search, and web research — via stdio transport.
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      MCP server that gives AI agents a ranked, evidence-backed board of what's trending in AI — repos, models, papers, news — and a daily brief, using fused signals from GitHub, Hacker News, npm, and more.
      132 PyPI
      4
      Apache 2.0