Nextcloud ChatGPT Bridge

Nextcloud ChatGPT Bridge

Enables ChatGPT and Codex to securely manage files in Nextcloud via OAuth-protected MCP tools, with tenant isolation and root-bound access.

Category
访问服务器

README

Nextcloud for ChatGPT & Codex

Nextcloud bridge mark

An open-source, tenant-isolated bridge that lets ChatGPT and Codex work with files in a user-owned Nextcloud workspace. The bridge keeps its OAuth identity separate from Nextcloud credentials and uses standards-based providers when native Nextcloud MCP capabilities are absent.

Product website: nextcloud-for-chatgpt.v4t0r.chatgpt.site

Landing page preview

Release status

v0.3.0 is the public-app and production-deployment release candidate.

The repository contains the application-side pieces required for a universal OAuth-protected MCP service:

  • request-scoped bridge identity and pseudonymous tenant isolation
  • Nextcloud Login Flow v2 without collecting a user's Nextcloud password
  • tenant-scoped metadata plus encrypted PostgreSQL credential storage
  • root-bound WebDAV and OCS providers with explicit read/write tools
  • conservative household invoice review that never approves, books, pays, or transmits
  • rate limiting, request limits, security headers, trusted-host checks, and stateless MCP transport
  • a DNS-rebinding-resistant HTTPS egress proxy and network-isolated production composition
  • database migrations, readiness checks, maintenance cleanup, and release preflight
  • a Codex plugin package plus reviewer-ready OpenAI listing and test material
  • a production landing site with privacy, terms, support, and security pages

The source release does not claim an active public service or OpenAI approval. Public availability still requires an operator-controlled production domain, external OAuth configuration, reviewer fixture, verified publisher identity, and successful OpenAI review.

The MCP bridge itself works without directory publication. Codex CLI, ChatGPT desktop, and the Codex IDE extension can connect directly; compatible remote MCP hosts can use a deployed HTTPS endpoint. ChatGPT web requires developer-mode access for an unpublished connection or installation of a published plugin. See deployment modes.

Real Nextcloud validation

The standards-based fallback path passed a live integration run against Nextcloud 33.0.7:

  • authenticated OCS discovery
  • WebDAV listing, create, upload, download, metadata lookup, move, and cleanup
  • user-visible app inventory and root-bound share inventory
  • isolated synthetic household workspace and invoice-review flow

Native Nextcloud Context Agent MCP was unavailable on that instance. The successful fallback run is therefore direct evidence that native MCP is optional rather than a hidden dependency. No credentials or instance-identifying values are recorded. See the compatibility matrix.

Architecture

ChatGPT / Codex
       |
       | OAuth access token
       v
Public stateless MCP boundary
       |
       | verified issuer + subject + scopes
       v
BridgeSessionContext
       |
       | pseudonymous tenant scope
       v
ConnectionService
       |-- PostgreSQL connection metadata
       `-- AES-256-GCM credential store
       |
       | Nextcloud Login Flow v2 credential
       v
Root-bound provider core
       |-- WebDAV files
       |-- OCS capabilities, apps, shares, revocation
       `-- Native Nextcloud MCP discovery when available

The bridge identity never contains a Nextcloud username or credential. Every connection, pending flow, household profile, and credential operation is scoped by the verified tenant context.

Public MCP tools

Read and discovery

  • get_nextcloud_capabilities
  • get_nextcloud_app_accesses
  • probe_native_nextcloud_mcp
  • list_files, search_files, list_nextcloud_shares
  • get_file_info, read_text_file, download_file_base64
  • list_nextcloud_connections, list_household_accounts, list_household_invoices
  • review_household_invoice

Create or modify private Nextcloud state

  • write_text_file, upload_file_base64, create_folder, move_file, delete_file
  • begin_nextcloud_connection, poll_nextcloud_connection, set_nextcloud_root
  • disconnect_nextcloud
  • configure_household_account, prepare_household_workspace
  • save_household_invoice_review

Every hosted tool has an explicit title, description, input schema, output schema, and risk annotations. The production tool contract is locked by automated tests and documented in submission/TOOL_ANNOTATIONS.md.

Security defaults

  • account root / and parent traversal are rejected before provider access
  • every returned WebDAV path is rechecked against the configured root
  • HTTPS and TLS verification are mandatory for hosted Nextcloud targets
  • public targets are resolved, validated, and IP-pinned by a CONNECT-only egress proxy
  • bearer tokens, app passwords, share tokens, raw invoice text, and full IBANs are excluded from model-visible results
  • credential ciphertext is tenant-bound through AES-GCM authenticated data
  • overwrite is opt-in; the configured root cannot be deleted
  • request body and transfer sizes are bounded
  • write, overwrite, move, delete, disconnect, and credential-revocation risks are explicit
  • hosted access is stateless and derives a fresh identity for every MCP request

Read SECURITY.md before deployment. Report vulnerabilities through GitHub private vulnerability reporting, never through a public issue containing secrets or private data.

Local development

Requires Python 3.11 or newer.

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e ".[dev,hosted]"
ruff check .
pytest

Copy .env.example to .env only for local testing. Use a dedicated Nextcloud account and app password, keep the root narrow, and never commit the resulting file.

Run the local stdio server:

nextcloud-chatgpt-bridge

Run sanitized read-only diagnostics or the explicit temporary write/cleanup smoke test:

nextcloud-chatgpt-diagnose
nextcloud-chatgpt-diagnose --write-test

Production deployment

The reference composition is in deploy/compose.production.yml. It separates PostgreSQL, the bridge, maintenance, egress, and TLS termination, and keeps the bridge off the direct external network.

cp deploy/.env.production.example deploy/.env.production
docker compose --env-file deploy/.env.production \
  -f deploy/compose.production.yml config
docker compose --env-file deploy/.env.production \
  -f deploy/compose.production.yml up -d

Supply database and encryption secrets as files outside Git. Apply migrations through the dedicated migrate service, then run nextcloud-chatgpt-preflight against the exact public MCP, OAuth, website, support, privacy, and terms URLs before any reviewer access.

See docs/PRODUCTION_DEPLOYMENT.md and docs/HOSTED_ACCEPTANCE.md.

OpenAI submission package

Repository-side review material lives in submission/:

  • canonical listing copy and starter prompts
  • positive and negative reviewer cases
  • reviewer fixture and runbook
  • exact tool-annotation inventory
  • release notes and final operational checklist

The final OpenAI submission remains an owner-controlled action. See docs/PLUGIN_SUBMISSION.md.

Project documents

License

Licensed under the Apache License 2.0. This independent project is not affiliated with or endorsed by Nextcloud GmbH or OpenAI.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选