Head-to-head, benchmarked

Seekstone vs obsidian-mcp-server.

obsidian-mcp-server is the most-downloaded Obsidian MCP server and the strongest REST-proxy option we benchmarked — its payloads stay bounded where other proxies balloon. Seekstone's difference is structural: it reads the vault from disk, so there's no Local REST API plugin to install, no running Obsidian app to depend on, ~118× faster warm search, and ~23× smaller search payloads on the 10,000-note benchmark vault.

Every number below comes from an open-source harness run on committed 1,000 / 5,000 / 10,000-note vaults — reproducible, not estimated. Full methodology on the benchmarks page; obsidian-mcp-server lives at cyanheads/obsidian-mcp-server.

At a glance

The comparison in one table.

Seekstone obsidian-mcp-server
Architecture Filesystem-direct, warm in-process index REST proxy via Local REST API plugin
Local REST API plugin Not needed Required
Obsidian app running Not needed — works with Obsidian closed Required
Search payload @ 10k notes (mean) 2.0 KB 47 KB
Warm search latency @ 10k notes 6.2 ms 732 ms (~118× slower)
Structured frontmatter queries Built-in (query_notes) — property/date/size predicates, ~350 B answers JSONLogic via REST
Scaling

What happens as your vault grows.

A vault only gets bigger. Here's both servers on the same committed vaults at 1,000, 5,000, and 10,000 notes — payload first, then warm latency.

Search payload — per query lower is better

Server1k notes5k notes10k notes
Seekstone 1.6 KB1.8 KB2.0 KB
obsidian-mcp-server 55 KB47 KB47 KB

Warm search latency lower is better

Server1k notes5k notes10k notes
Seekstone 1.1 ms3.1 ms6.2 ms
obsidian-mcp-server 82 ms356 ms732 ms
The architecture

Why the numbers differ.

obsidian-mcp-server is the REST proxy done well: it caps what it returns, so the payload story is bounded rather than catastrophic. The remaining ~23× payload gap and the dependency chain are structural — it can only answer while Obsidian is open with the Local REST API plugin installed, configured with an API key, and reachable over local HTTPS. Seekstone reads the vault directly from disk: nothing to install inside Obsidian, nothing that has to be running.

Latency is where the round-trips show: 732 ms warm search at 10,000 notes vs 6.2 ms from Seekstone's warm in-process index — and its latency grows ~9× from a 1k-note vault to a 10k-note vault, while Seekstone's barely moves. For an agent making many tool calls per task, milliseconds-vs-seconds compounds quickly.

Honest accounting

Where obsidian-mcp-server has the edge.

No tool wins every axis, and a comparison that pretends otherwise isn't worth your trust.

The most-downloaded option.

obsidian-mcp-server is #1 by npm downloads among Obsidian MCP servers — a real signal of maturity, documentation, and community testing that a newer server has to earn.

Bounded payloads for a REST proxy.

Unlike other REST proxies, its search responses hold roughly flat (~47 KB) as the vault grows — the best payload discipline of any proxy we benchmarked.

Leans on Obsidian's plugin ecosystem.

Running inside the app's REST API lets it integrate app-side capabilities — such as Omnisearch-backed relevance ranking — that a filesystem-direct server has to reimplement itself.

Don't take our word for it

Reproduce every figure.

The harness, the query set, and all three vaults are committed to the Seekstone repo. Nothing on this page is a marketing estimate.

Terminal
git clone https://github.com/shaqmughal/seekstone cd seekstone && npm install npm run harness -- bench --backend fs \ --vault packages/harness/fixtures/vault
Switching

Trying Seekstone takes about 30 seconds.

Seekstone is a standard MCP stdio server — it works in any client that runs obsidian-mcp-server, side by side with your current setup, and doesn't touch your notes until you ask it to. Guided setup below, or see the per-client install guides.

Terminal
npx -y seekstone init
FAQ

Common questions.

Is Seekstone a good alternative to obsidian-mcp-server?

If you want vault access without installing the Local REST API plugin or keeping Obsidian running, yes — that's the structural difference. On the 10,000-note benchmark vault Seekstone also returns ~23× smaller search payloads (2 KB vs 47 KB) and answers ~118× faster warm (6.2 ms vs 732 ms). Both are standard MCP stdio servers, so switching is a config change.

What does obsidian-mcp-server do better?

It's the most-downloaded Obsidian MCP server, it has the best payload discipline of any REST proxy we benchmarked, and running inside the app lets it use app-side features like Omnisearch-backed relevance ranking. If your workflow depends on plugin-side capabilities, that's a genuine advantage.

Do both servers edit notes safely?

Seekstone's write path is verified byte-for-byte by its open-source harness: frontmatter is edited in place with key order, quote style, and comments preserved. We haven't benchmarked competitors' write safety, so we make no claim about obsidian-mcp-server there — the harness is public if you want to run that comparison.