io.github.meemoprasad由 GitHub 帳號 meemoprasad 發布v1.0.1更新於 Oct 24, 2026
Repository knowledge graph MCP server for codebase understanding and debugging.
發布方自己的服務：BrainDog
一個遠端 MCP 伺服器，為 AI 助理提供已索引程式碼庫的結構化脈絡：符號、呼叫端、被呼叫端與相依關係。
Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。
閱讀 https://sourceweft.com/mcp/io.github.meemoprasad%2Fmeeba-brain/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。把它新增到你客戶端的 mcpServers 設定中。
{
"mcpServers": {
"meeba-brain": {
"type": "http",
"url": "https://meeba-context-mcp-469248886325.us-central1.run.app/mcp"
}
}
}Give your AI coding assistant a structural understanding of your codebase.
Meeba Brain MCP connects Cursor to Meeba Brain's repository knowledge graph, helping it understand how functions, files, services, and dependencies are connected.
Instead of relying only on text search, Cursor can use Meeba Brain's structural context to investigate unfamiliar code, trace relationships, and reason about changes with better grounding.
Text search helps you find where a name or phrase appears.
A knowledge graph helps you understand how the code is connected.
For example, instead of only finding refundOrder, you can investigate:
The difference is not simply finding more text. It is understanding the relationships between the code.
When debugging, the file showing the error is not always the file causing the problem.
A bug may involve several connected components:
API route → service → worker → dependency
Without understanding those relationships, an AI assistant may suggest a plausible fix that changes the wrong layer or misses an affected caller.
Meeba Brain provides Cursor with repository-grounded structural context so Cursor can investigate the relevant relationships before proposing a change.
This helps Cursor identify the relevant code, understand potential impact, and propose better-informed fixes.
Meeba Brain does not guarantee that every fix is correct. Its purpose is to give Cursor better evidence and structural context for reaching the right fix.
Meeba Brain MCP provides access to repository knowledge graph capabilities, including:
Meeba Brain MCP requires a connected, indexed repository.
The repository is connected and indexed on Meeba Brain, not inside Cursor.
Go to:
After signing in, connect the repository you want to investigate on Meeba Brain.
Meeba Brain needs to build the repository's structural knowledge graph.
Wait until indexing is complete before connecting the MCP to Cursor.
Open the MCP connection page:
https://meeba.ai/brain/connect
Enable MCP and copy the key provided by Meeba Brain.
Keep this key private.
In Cursor:
Use the exact server URL and authentication header shown in Meeba Brain.
After saving the configuration, open the MCP settings and make sure Meeba Brain is enabled.
Open your repository in Cursor and ask a question such as:
"Which services call refundOrder()?"
"Explain how this feature flows through the codebase."
"What could be affected if I change this API?"
"Trace the code flow from this route to the implementation."
"Why is this function being called?"
Start with one real function, dependency, or bug you want to understand.
Meeba Brain MCP is not a replacement for Cursor.
It is a repository knowledge graph integration that gives Cursor additional structural context.
Cursor remains responsible for the coding conversation, reasoning, and code changes.
Meeba Brain provides the repository intelligence that helps make that reasoning more grounded.
Learn about Meeba Brain:
Connect your repository and enable MCP:
https://meeba.ai/brain/connect
Understand the relationships. Investigate the evidence. Make better-informed changes.
來源：README.md，提交 406680a
Hybrid semantic + keyword search over the connected repository's actual indexed content. Use this for open-ended questions ("where is rate limiting implemented?", "how are uploads validated?") where you don't already know an exact symbol or file name -- for those, use get_symbol_definition or get_dependencies instead, they're more precise.
query必填top_kFinds where a function, class, or method is actually defined in this repo -- exact file, line range, and signature-relevant metadata (async, decorators, docstring if present). Use this before assuming a symbol's location/signature from memory or from a similarly-named symbol in a different file.
name必填Every place in this repo that calls the given function or method -- a COMPLETE list from the parsed call graph, not a best-effort guess. Use this before renaming/changing a function's signature, or when you need to understand a function's blast radius before modifying it.
function_name必填Every function/method the given function calls -- what it depends on to do its job. Use this to understand a function's behavior by seeing what it delegates to, before assuming what it does from its name alone.
function_name必填What a specific file imports/depends on. Use this to understand a file's place in the codebase before assuming what's available to it, or before moving/refactoring it.
file_path必填已滿足適用檢查項的 90%
該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。
請在儲存庫或註冊中心條目中宣告授權條款。
dog.brain官方發布者
BrainDog 是一個託管記錄庫，集中保存專案、決策與金錢，接入的 AI 代理在開工前讀取、結束後寫入。
由 GitHub 帳號 Nomannazir 發布
為本機或公開 git 程式碼庫建立索引，讓助理能檢索程式碼、追溯 git 歷史、分析變更影響並找出過時文件，結果都附上檔案路徑與行號引用。
由 GitHub 帳號 zhang-liz 發布
為 AI 助理提供職場人脈記憶，資料以圖的形式存放在 FalkorDB，可查詢聯絡人、依主題找人並記錄你見過的人。
由 GitHub 帳號 ara-5 發布
面向AI代理程式的共享Graph-RAG記憶：markdown vault、wikilink召回、任務、訊息傳遞。AI 翻譯
由 GitHub 帳號 make0-official 發布
本機離線的記憶伺服器，讓 Claude Code、Cursor、Codex 等 MCP 代理共用同一份具版本紀錄的長期事實庫。