seekstone: local MCP server for fast Obsidian vault access
seekstone, by Shaqmughal, is an MCP server that gives AI models direct access to Obsidian vaults to reduce context overhead and speed retrieval. The tool reads vault files from disk, exposes a suite of search and note-management tools, and supports keyword plus semantic search. It targets Obsidian users and developers who need low-latency, model-facing access to personal knowledge bases for tighter model-assisted workflows.
Designed to supply concise excerpts so models keep reasoning tokens
The tool purposely returns ranked excerpts instead of full-note dumps, typically around 2 KB, so the model's context window remains available for reasoning. Its 17 specialized tools cover searching, reading, writing, and note management; examples include patching frontmatter and handling backlinks. The app bypasses the Obsidian application by reading the filesystem directly, which changes the retrieval step from bulk transfer to targeted context delivery.
Delivers millisecond search latency and large payload reductions
Seekstone reports single-digit millisecond keyword search times and claims up to 500x speed improvement over REST-proxy alternatives. It also advertises payload reductions measured as much as 47,000x compared to full-note retrieval, which materially lowers the token count sent to models. Semantic search runs locally using on-device embedding models, so meaning-based queries do not require remote lookups.
Integrates with MCP hosts but has specific install paths
The app is compatible with MCP-compliant hosts such as Claude Desktop, Cursor, VS Code, and others, and offers an automated CLI setup via seekstone init. CLI installation requires Node.js version 22 or newer, while a one-click .mcpb bundle exists for Claude Desktop with no prerequisites. The developer emphasizes zero-dependency operation with no Obsidian plugins required.
Keeps vault data on the local machine and emphasizes privacy controls
Seekstone runs entirely locally and does not send vault files to external servers; the vault data stays on the user's machine. The developer positions the project as privacy-first and local-only, and the tool's architecture avoids relying on Obsidian's APIs. This design suits workflows where retaining file ownership and avoiding cloud transfer are priorities.
Practical choice for users who prioritise local, model-accessible KBs
seekstone is a practical option for Obsidian users and developers who need fast, context-conscious model access to their notes. Expect the strongest benefit when you design prompts to exploit brief, ranked excerpts and when you verify critical facts independently. For workflows that value local data handling and model reasoning over full-document ingestion, the tool is a suitable integration point.
