Obsidian GitHub MCP
A Model Context Protocol server that gives AI agents bounded, auditable access to propose changes to an Obsidian vault stored in GitHub, with per-agent expiring tokens, read/write/deny path policies, mutability rules, frontmatter checks, and human-reviewed proposal branches.
README
Obsidian GitHub MCP
A white-label Model Context Protocol server that gives selected AI agents bounded, auditable access to propose changes to an Obsidian vault stored in GitHub.
Each agent receives its own expiring MCP bearer token. The GitHub credential remains server-side. Read and write paths, denials, protected surfaces, proposal branches, and token revocation are controlled by operator-owned YAML files.
This is an independent open-source project. It is not affiliated with or endorsed by Obsidian or GitHub.
Why use it?
Giving an agent a general GitHub token is difficult to constrain and easy to misuse. This server creates a narrower boundary:
AI agent + individual MCP token
|
v
identity + fixed broker
|
v
agent write scope AND broker publication scope
|
v
deny, path, frontmatter, mutability, size, and secret checks
|
v
server-side GitHub App token
|
v
proposal branch for human review
The service has no tools for deleting files, merging branches, pushing to main or master, changing repository settings, editing workflows, or exposing credentials.
Features
- Stateless MCP Streamable HTTP transport
- Unique, expiring bearer tokens stored as SHA-256 bindings
- Per-agent read, write, and deny path globs
- Separate Git broker allowlists and fixed proposal branches
- Configurable frontmatter requirements for shared or sensitive surfaces
living,review-first,append-only, andimmutablenote handling- Path-traversal prevention and file-extension/size allowlists
- Credential-shaped-content detection
- Repository-restricted GitHub App tokens with PAT fallback for local testing
- Content-free JSONL audit records
- Host/origin validation, security headers, rate limiting, and HTTP timeouts
- Docker, CI, tests, and deployment documentation
MCP tools
| Tool | Purpose |
|---|---|
vault_whoami |
Show the authenticated identity, broker, branch, and policy scopes. |
vault_list_allowed_paths |
Explain effective read/write/deny boundaries. |
vault_explain_access |
Dry run a single path. Report whether read, list, or write would be permitted and which policy rule decides it. No GitHub call. |
vault_list_files |
List an authorized directory while filtering denied children. |
vault_read_file |
Read an authorized UTF-8 vault file. |
vault_create_file |
Create a file on the identity's fixed proposal branch. |
vault_update_file |
Update a permitted file while respecting mutability. |
vault_append_file |
Append while preserving all existing bytes. |
Requirements
- Node.js 22+
- An Obsidian vault in a GitHub repository
- A dedicated non-default proposal branch for each broker
- A GitHub App installed only on the vault repository with
Contents: read and write, or a fine-grained PAT for local testing
Quick start
git clone https://github.com/YOUR_ACCOUNT/obsidian-github-mcp.git
cd obsidian-github-mcp
npm ci
cp .env.example .env
cp config/policy.example.yaml config/policy.yaml
cp config/tokens.example.yaml config/tokens.yaml
npm run token:new
npm run check
npm start
Configure .env, replace the sample policy, and put only the generated token's SHA-256 in config/tokens.yaml. Give the raw bearer token to exactly one MCP client.
The default endpoints are:
GET http://127.0.0.1:3210/healthz
POST http://127.0.0.1:3210/mcp
MCP clients must send:
Authorization: Bearer vault_<generated-token>
Permission model
The effective write scope is the intersection of the agent grant and its broker:
policy_version: 1
authority: operator-only
agents:
- id: example-writer
name: Example Writer
role: bounded knowledge worker
disclosure_ceiling: internal
entry_reads: [README.md]
read: [Knowledge/**, Workspace/**]
write: [Workspace/Shared/**, Workspace/Example Writer/**]
deny: [Private/**, People/**, .obsidian/**]
git_brokers:
- broker: example-writer
branch: agent/example-writer
allow: [Workspace/Shared/**, Workspace/Example Writer/**]
never_versioned: [Private/**, People/**, .obsidian/**, secrets/**]
surface_rules:
- path: Workspace/Shared/**
required_frontmatter:
disclosure: public-safe
An auth binding then maps one token hash to that identity and broker. The binding cannot grant paths that the policy does not grant.
See docs/PERMISSIONS.md for provisioning and revocation.
GitHub authentication
Production deployments should use a GitHub App installed only on the vault repository. Configure:
GITHUB_APP_ID=123456
GITHUB_APP_INSTALLATION_ID=12345678
GITHUB_APP_PRIVATE_KEY_FILE=/run/secrets/vault-mcp.pem
The server creates short-lived installation tokens restricted again to the configured repository and contents: write. A repository-scoped GITHUB_TOKEN is supported as a local-development fallback.
Deployment
The included container runs as a non-root user with a read-only root filesystem. Put TLS, a private overlay network, or an authenticated gateway in front of the service before making it remotely reachable.
See docs/DEPLOYMENT.md.
Verification
npm run check
npm audit --audit-level=high
Tests cover policy parsing, write-scope intersection, deny precedence, traversal attempts, frontmatter rules, mutability, secret detection, token bindings, and fixed-branch GitHub writes.
Security
Read SECURITY.md before deployment. Report suspected vulnerabilities privately rather than opening a public issue with exploit details.
License
Apache-2.0. See LICENSE.
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.