Contexter

io.github.maaakso由 GitHub 帳號 maaakso 發布v1.0.0更新於 Oct 28, 2026

Share one project context across ChatGPT, Claude, Telegram and any MCP client.

已驗證Streamable HTTP可網頁執行知識與記憶溝通與協作

概覽

AI 產生的概覽

託管式遠端 MCP 伺服器,讓 ChatGPT、Claude、Telegram 等 MCP 用戶端共用同一份專案脈絡。

可以做什麼
  • 讀取脈絡、脈絡卡片、動態串、紀錄、來源、產出物、檔案與圖片
  • 儲存與更新產出物、檔案、脈絡與紀錄,並設定脈絡通知
  • 分享脈絡、檢查分享對象、把內容投遞到聊天、邀請房間加入脈絡
  • 移除、還原、刪除或退出紀錄與脈絡,並撤銷已分享的存取權
適用情境
適合在多個 AI 助理之間或與他人協作同一專案,希望用一份共享的已儲存資料取代在每個工具裡重複說明背景。也適合想把群組聊天中的決定留下來、再由各成員在自己的 AI 中接續的 Telegram 群組。
執行需求
需要遠端 MCP 端點(Streamable HTTP),不必安裝,也沒有 API 金鑰。必須以 OAuth 2.1 登入,並經 Telegram 核准一次。可用於 Claude、ChatGPT,以及任何支援 Streamable HTTP 與 OAuth 2.1 的用戶端。
安裝前請注意
脈絡可分享給他人與 Telegram 群組,儲存的紀錄可能對這些對象可見,儲存前應確認可見範圍。寫入類工具會建立、更新、分享與刪除資料,刪除僅在 7 天內可還原。在 Telegram 群組中,已加入者的訊息會被收集為紀錄。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Contexter,將其新增到工作區。
  2. 連接 所需的 憑證。它們會被私密儲存,絕不會在公開頁面上展示。
  3. 為需要使用其工具的對話啟用該服務。

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。

交給代理程式安裝

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.maaakso%2Fcontexter/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。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。 客戶端會在首次連接時完成 OAuth 登入。

{
  "mcpServers": {
    "contexter": {
      "type": "http",
      "url": "https://contexterai.com/mcp"
    }
  }
}

README

Contexter

Share one AI context across ChatGPT, Claude and Telegram.

Save what you worked out with one AI, continue it in another — and share the context with people in Telegram, who keep working on it in their own AI.

Contexter is a hosted remote MCP server. This repository is its public documentation: the connection details, the tool catalogue and what does and does not move between tools. The server itself is hosted and closed-source — there is nothing to install and no key to manage.

  • Homepage: https://contexterai.com
  • MCP endpoint: https://contexterai.com/mcp
  • Transport: Streamable HTTP
  • Authentication: OAuth 2.1 (with protected-resource metadata discovery)
  • Telegram bot: @ai_contexter_bot
  • Status: early access, free, open to everyone

The problem it solves

You plan something in ChatGPT. You continue it in Claude. You discuss it in a Telegram group. Every hop, you re-explain the project — and the person you are working with cannot pick it up at all, because your history lives inside your account.

Contexter gives the project one context instead of one copy per tool. Both assistants read and write the same records. You can share that context with a person, and they continue it in whichever AI they use.

Nothing is imported automatically. Connecting an assistant does not hand Contexter your chat history. Only the records you explicitly save — or ask your assistant to save — end up in a context.


Connecting

Claude

Add a custom connector pointing at https://contexterai.com/mcp. Claude discovers the OAuth endpoints on its own; you sign in through Telegram and approve access once.

ChatGPT

Add Contexter as a connector with the same URL. Same OAuth flow.

Any other MCP client

Anything that speaks Streamable HTTP with OAuth 2.1 works the same way — point it at https://contexterai.com/mcp.

Full walkthroughs, including what travels between each pair of tools:


What you can ask an assistant to do

Once connected, you talk to your assistant normally. For example:

Create a context called "Kitchen renovation" in Contexter and put my measurement notes in it.

What is in my Dolomites hike context?

Save this comparison to the AI-bot context and share it with Anna.


Tool catalogue

Twenty-six tools in four groups. Write tools require a human decision — the client acts for a person, and irreversible steps (sharing, publishing to a room, deletion) are confirmed rather than assumed.

Reading (8) — list_contexts, get_context_card, get_context_feed, list_records, list_sources, get_artifact, get_file, read_image_content

Saving (7) — save_artifact, save_file, create_context, add_records_to_context, update_artifact, update_context, set_context_notifications

Sharing (5) — share_context, revoke_context_access, check_share_audience, deliver_to_chat, invite_room_to_context

Removing (6) — remove_record_from_context, delete_record, leave_context, restore_record, delete_context, restore_context

Scopes, response shapes and the reasoning behind each one: contexterai.com/developers.


Use cases

Two assistants, one project. Research in ChatGPT, drafting in Claude. The plan lives in a context both reach, so neither one starts from an empty page.

Two people, two different AIs. You share the context, not a document export. The other person opens it from whatever assistant they already pay for.

A group chat that does not lose its decisions. Add the bot to a Telegram group. Messages from the people who opt in become records; the context built from the chat is shared with those same people, and each of them works on it in their own AI, outside the group.


Privacy

  • AI chat history is never imported. Saving is always explicit.
  • A record's visibility is the visibility of the context it is in. There are no hidden audiences: every shared context shows exactly who can see it.
  • In a Telegram group, only the messages of people who joined in are collected. Anyone who has not is not collected at all.
  • Deletion is reversible for 7 days.

Privacy policy · Security model in plain words · Terms · Subprocessors


Support

Open an issue here, or message @ai_contexter_bot in Telegram. Questions and answers for people already using it: contexterai.com/help.

來源:README.md,提交 4324aea

工具

0
該伺服器需要登入,因此 SourceWeft 無法在沒有帳號的情況下讀取其工具,也未收錄。已連接帳號所能看到的內容可能因帳號及其權限而異。

版本歷史

1
  1. v1.0.0最新Sep 16, 2026

評分

已滿足適用檢查項的 75%

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。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分

    請在儲存庫或註冊中心條目中宣告授權條款。