Unobsidian — Obsidian Vault Knowledge Graph

    1

    Explore local Obsidian note links, backlinks, and tags with your AI agent. Free and open source.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Unobsidian — Obsidian Vault Knowledge Graph

    Example session with this skill installed

    In a small example vault, show the one-step neighborhood of ProjectA as D3 JSON. Preserve incoming links, outgoing links, and embed relations. Explain which notes are direct neighbors.

    • Read your context and instructions
    • Compiled the unobsidian

    Example: one-step neighborhood

    This reproducible demonstration uses synthetic notes, not a personal vault:

    • ProjectA.md contains [[Plan]] and #project.
    • Plan.md contains [[ProjectA]], ![[ProjectA]], and [[Research]].
    • Research.md contains #research.

    Command (replace the example vault path with your own):

    python "${CLAUDE_SKILL_DIR}/scripts/get_neighbors.py" --vault "/path/to/example-vault" --node "ProjectA" --depth 1 --format d3
    

    Complete JSON returned by Unobsidian 0.1.1 for this example:

    {
      "nodes": [
        {"id": "Plan", "type": "file"},
        {"id": "ProjectA", "type": "file"},
        {"id": "#project", "type": "tag"}
      ],
      "links": [
        {"source": "Plan", "target": "ProjectA", "relation": "wikilink"},
        {"source": "Plan", "target": "ProjectA", "relation": "embed"},
        {"source": "ProjectA", "target": "Plan", "relation": "wikilink"},
        {"source": "ProjectA", "target": "#project", "relation": "tag"}
      ]
    }
    

    Plan is a direct note neighbor: it both links to and embeds ProjectA, while ProjectA links back to Plan. #project is a direct tag neighbor. Research is two steps away, so it is excluded at depth 1. No note contents were changed.

    Next step: use depth 2 to include Research, or resolve Plan to its path and read the actual note before summarizing its contents.

    ProjectA.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Explore your vault with an AI agent

    Unobsidian is a free, MIT-licensed skill for exploring a local Obsidian-style vault as an explicit knowledge graph. Obsidian does not need to be installed or running.

    What it does

    • Extracts note links, embeds, and tags into compact adjacency text or D3-compatible JSON.
    • Explores 1–3 steps around a note or tag while preserving original link directions and relationship types.
    • Finds backlinks and searches nested tags.
    • Resolves note IDs to local file paths so your agent can read the relevant notes.
    • Generates an interactive HTML graph visualization on request.

    Requirements and scope

    Requires Python 3.10+, obsidian-parse==0.1.1, local access to your vault, and an agent able to run the bundled Python scripts. Designed for Claude Code; other agents need equivalent local script access. A text-only chat or marketplace connector alone cannot execute the scripts.

    Graph queries do not edit notes. Visualization writes an HTML file and loads D3.js from d3js.org. Same-name notes in different folders are not distinct graph nodes. Queries reparse the vault each time; this is not semantic search or a live synchronized index. Read note contents before drawing conclusions from graph links.

    Free and open source

    Source, installation instructions, releases, and issue tracker: https://github.com/agent-hanju/unobsidian

    If you find Unobsidian useful, a star on the original GitHub repository is appreciated — entirely optional.

    Questions or feedback: agent.hanju@gmail.com. Bug reports are also welcome on GitHub Issues.

    Version 0.1.1. Independent open-source project; not affiliated with Obsidian.

    Requires local filesystem and Python execution access; not usable in a text-only chat without the bundled scripts. Same-name notes in different folders share a graph ID. Queries reparse the vault on each call and do not perform semantic search or synchronization. BFS depth is 1–3. Optional HTML visualization loads D3.js from d3js.org and writes its output file.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 29 days ago

    • Free with an Agensi account

    Requires

    Python 3.10+
    obsidian-parse 0.1.1
    Obsidian-style vault folder
    Terminal access
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Browser
    Network Access

    Reads the user-selected local vault (.md, .canvas, .base and vault ignore settings). Python commands run locally. Graph and tag queries do not edit notes. Optional visualization creates/replaces only its specified HTML output and may open it in a browser; the HTML fetches D3.js from d3js.org. Installing requirements downloads packages from PyPI. No account connection, API key, or remote vault upload is required by the scripts.

    Listed29 days ago

    What's inside

    Frequently Asked Questions

    Yes. Unobsidian is free and MIT licensed. Source, releases, and installation instructions are at https://github.com/agent-hanju/unobsidian. If it helps you, an optional GitHub star is appreciated.

    Python 3.10+, the included requirements.txt, a local vault, and an agent able to execute the bundled scripts and read local files. Designed for Claude Code; other agents need equivalent local access. A text-only chat or marketplace connector alone is insufficient. Obsidian does not have to be running.

    Email agent.hanju@gmail.com for questions or feedback. For reproducible bugs, please open an issue at https://github.com/agent-hanju/unobsidian/issues. Do not include private vault contents in public reports.

    New in Productivity