discoveree-desktop
Provides a local, agent-maintained product context layer over MCP, serving structured strategy, competitor, customer feedback, and feature inventory data to any AI tool such as Claude, Cursor, or ChatGPT, with local data storage, bring-your-own-LLM support, and weekly evidence-based roadmap reviews that can sync suggestions to Jira or Linear.
README
Discoveree Desktop
A local, agent-maintained context layer for your product, served over MCP to any AI tool.
Discoveree keeps a structured, always-fresh picture of your product world — strategy, competitors, customers and feedback, and your product's own feature inventory — maintained by agents and served over the Model Context Protocol to whatever AI tools you already use: Claude, Cursor, ChatGPT, or your own agents.
Status: pre-alpha. The desktop edition is under active development and not yet ready for use. The build plan lives in docs/build-brief.md.
Why
AI tools are only as good as the context they're given. Prose documents rewritten by prompts silently accumulate duplicates and drift. Discoveree holds product context as a typed, validated schema — stable IDs, provenance, confidence, freshness accounting — kept current by deterministic pipelines (hash-diff changelog monitoring, help-centre crawling, review mining), and serves it to every tool and teammate from one source of truth.
- Your data stays local. Competitive and roadmap data never leaves your machine.
- Bring your own LLM keys. One key from any major provider is enough.
- Any AI tool connects. Local MCP over stdio and localhost HTTP.
- Judgment, not just storage. A weekly, evidence-cited review of whether your roadmap is building the most valuable things — with suggestions you can accept into Jira or Linear.
Structure
client/ React SPA (shared by desktop and team deployments)
server/ Express server — embedded in the desktop app, deployable shared for teams
shared/ Schema and types (Drizzle + Zod)
docs/ Build brief and design docs
One codebase, two deployments: the desktop app embeds the server with a local database; the team tier deploys the same server shared. Local/team is a deployment target, not a fork.
Licence
Functional Source License, Version 1.1, ALv2 Future License (FSL-1.1-ALv2): source-available — you can read, run, and modify the code for any non-competing purpose, and each version becomes Apache 2.0 licensed two years after release.
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。