vault-mcp-bridge
Read-only MCP bridge that exposes secure search and fetch tools over an Obsidian-compatible Markdown vault, enabling ChatGPT to query notes without write access.
README
Vault Bridge
Read-only access to a private Obsidian vault from ChatGPT.
Vault Bridge is a macOS application that copies a filtered snapshot of a vault to an isolated Docker project on your own Linux server. OpenAI Secure MCP Tunnel connects that private MCP server to ChatGPT without a public port or inbound access to the Mac.
Install with Codex · Releases · Architecture · Security
The current Apple Silicon preview is integrity-checked but not Developer ID signed or notarized. Do not bypass a macOS security warning for an untrusted download. The supported preview path is the Codex installation runbook.
What it does
- selects any local Obsidian vault;
- connects to an existing Linux server over SSH;
- installs an isolated, resource-bounded Docker Compose project;
- publishes complete vault snapshots on startup, every five minutes, after resume, or on demand;
- exposes only
searchandfetchto ChatGPT; - shows connection state, synchronization state, and an aggregate activity log in one desktop application.
- provides one redacted
doctor --jsoncheck for the vault, SSH/runtime, OpenAI tunnel, and last synchronization result.
The application does not write to the vault. It does not expose iCloud, a local filesystem, the Docker socket, a shell, or a public server port.
Install
For an agent-guided installation, open docs/install-with-codex.md, copy the prompt, and paste it into Codex. The runbook keeps account login, credentials, and the first SSH fingerprint approval with the owner.
Requirements:
- Apple Silicon Mac;
- Linux server reachable over SSH;
- Docker Engine with Docker Compose on the server;
- access to OpenAI Secure MCP Tunnels and the corresponding ChatGPT connection surface.
Release downloads include a DMG, ZIP, and SHA-256 manifest. Read the release notes before installing; preview binaries are not yet notarized.
Architecture
flowchart LR
phone["Obsidian on phone"] <-->|"iCloud or Obsidian Sync"| vault["Vault on Mac"]
vault -->|"local read-only scan"| app["Vault Bridge"]
app -->|"SSH snapshot upload"| runtime["Docker project on your server"]
runtime -->|"outbound HTTPS"| tunnel["OpenAI Secure MCP Tunnel"]
tunnel <--> chat["ChatGPT"]
The vault on the Mac remains canonical. The server copy is replaceable and activates as one complete generation after validation. Phone-to-Mac freshness remains the responsibility of iCloud or Obsidian Sync.
ChatGPT tools
The MCP surface is deliberately small:
search({ query })returns bounded matches with opaque IDs;fetch({ id })returns one document from the active snapshot.
There is no model-facing browse-all, write, delete, sync-control, raw-path, SQL, or shell tool. Vault content is treated as untrusted source data.
Secure MCP Tunnel supports private connections and developer-mode testing. It does not publish Vault Bridge to the public ChatGPT app catalog.
Security model
- Hidden directories,
.obsidian,.git,node_modules, and symlinks are excluded by default. - SSH host identity is approved once and pinned.
- The server process is non-root and resource-bounded, with no host ports, host mounts, privileged mode, or Docker socket.
- The OpenAI runtime key is accepted through the native UI or stdin and stored with macOS encrypted storage. It is not accepted in setup JSON or process arguments.
- Activity records counts, bytes, duration, trigger, generation, and a bounded component/error code when an operation fails. It does not record note contents, titles, queries, credentials, addresses, or local paths.
- The local journal retains at most 200 events. Container logs rotate at 8 MiB × 3 files, and temporary deployment secrets are removed after every setup attempt.
- The VPS administrator and Docker daemon can inspect the replica. Use a server you trust.
See the threat model and deployment contract. Report vulnerabilities through SECURITY.md, without vault evidence in a public issue.
Status
| Surface | Current preview |
|---|---|
| macOS | Apple Silicon; not notarized |
| Vault data | Markdown, Canvas, and Bases text |
| MCP | Read-only search and fetch |
| Sync | Startup, five-minute schedule, resume, manual |
| Runtime | Docker Compose; Linux amd64 and arm64 |
| Installation | Codex runbook or source build |
The public roadmap is tracked in GitHub Issues.
Development
Requires Node.js 24+ and pnpm 10.
pnpm install --frozen-lockfile
pnpm check
pnpm dev
Use only synthetic fixtures during development. The repository is a TypeScript monorepo:
apps/desktop— macOS application and deployment owner;apps/agent— vault scanner and snapshot publisher;apps/server— SQLite/FTS5 store and MCP server;deploy/secure-tunnel— isolated Docker Compose deployment;packages/*— shared contracts and libraries;apps/edge,deploy/runtime— advanced public HTTPS/OAuth mode.
See CONTRIBUTING.md. Licensed under Apache-2.0.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.