Hi all,
I’ve built an MCP server that lets an AI assistant (Claude, ChatGPT, local models, anything that supports MCP) answer questions about your tasks using the full Obsidian Tasks query language.
Disclaimer
- Is this plugin open source? Yes, MIT licensed.
- Is this plugin completely free? Yes
- Is this plugin vibe-coded beyond the author’s ability to comprehend how it works? I still understand how it works, but it’s more or less vibe-coded.
MCP (Model Context Protocol) is an open standard for giving an assistant tools it can call. Here the tool takes a Tasks query and runs it over your vault.
What makes it different from the other Tasks MCP servers I know: it doesn’t reimplement the query language. It runs the Tasks plugin’s own engine, unmodified, outside of Obsidian. So everything that works in a tasks code block works here: boolean combinations, relative dates, recurrence, is not blocked, grouping and sorting. Your global filter and custom statuses are respected. If the engine doesn’t understand a line, you get an error and an explanation, not a silently smaller result.
A few more things:
- Obsidian doesn’t need to be running.
- The vault is only ever read, never written.
- It only reads inside the vault folder you configure.
Example: “What do I have to get done this week? Leave out anything that’s still waiting on something else.” The assistant turns that into a query like not done / due before in 7 days / is not blocked and answers from the result.
It’s a first release (v0.1.0), and it’s not on npm yet, so you need Node.js 22+ and have to clone and build it. Instructions are in the README:
It’s not affiliated with Obsidian or the Tasks plugin. All credit for the engine goes to the Tasks authors.
I’d really appreciate feedback. Issues on GitHub or replies here are both welcome.