Practical overview
When it fits
Use Engraph when an AI assistant needs searchable, connected context from an Obsidian vault, or when a local application needs a vault-search interface. Choose MCP for agent-tool access and the optional HTTP REST API for scripts or web integrations.
Who it is for
- Obsidian users who want AI agents to search linked Markdown notes.
- Developers connecting a local application or script to vault search through HTTP REST.
- Agent builders who want note-reading, context-building or documented note-editing tools over a vault.
Documented uses
- Index a Markdown vault and retrieve relevant notes using hybrid search.
- Give an MCP-compatible agent structured access to vault search and note tools.
- Return topic or person context assembled from related notes and wikilinks.
- Expose vault operations to scripts or applications through the HTTP REST API.
- Create notes or make section-level edits through documented agent tools.
Requirements & limits
The repository identifies Engraph as open source under the MIT license. Listed installation routes include Homebrew for macOS, prebuilt binaries for macOS arm64 and Linux x86_64, and a Cargo source installation that requires CMake for llama.cpp. First indexing downloads an embedding model.
- Engraph is a self-hosted local tool, not a hosted vault-search service; its README describes local indexing and inference.
- The first index downloads an embedding model of about 300 MB, while an optional larger model supports intelligence features.
- MCP and HTTP are separate interfaces: MCP serves agent tools, and HTTP adds REST endpoints for scripts and web-based agents.
- HTTP setup includes API-key authentication and permissions; the README also documents a no-auth option restricted to 127.0.0.1 for local development.
- ChatGPT Actions require extra configuration, including HTTP server setup and a public tunnel; this is distinct from the local MCP workflow.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Index a vault and search its notes
Create a local index of an Obsidian vault and run a search against it.
Before you start
- Engraph installed using a repository-listed installation route.
- Path to an Obsidian Markdown vault.
- Network access on first indexing to download the embedding model (about 300 MB).
Steps
Run
engraph index ~/path/to/vault, replacing the example path with the vault location.Run
engraph search "how does the auth system work"with a question relevant to the notes.
Expected outcome
Engraph returns ranked note results; the repository example includes a result reached through wikilink graph expansion.
Guide limitations
- The first index downloads an embedding model; the README does not give a duration for indexing in general.
- The first-index download means this setup stage is not entirely offline.
Start the MCP server for an agent client
Expose an indexed vault through Engraph's MCP server for an MCP client.
Before you start
- Engraph installed.
- An indexed vault and its path available to Engraph.
- An MCP client that can be configured to launch a local command.
Steps
Configure the MCP client to launch
engraphwith the argumentserve; the README shows a Claude Code configuration example.Start or reconnect the MCP client using that configuration.
Expected outcome
The client can access the MCP tools exposed by `engraph serve`, including search and note-reading tools.
Guide limitations
- The README gives a Claude Code example but does not document configuration details for every MCP client.
- The repository describes available tools; this guide does not verify a client connection.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- devwhodevs/engraph · Official homepage title · retained in its source language