SkillAgentSearch skills...

claude-obsidian

Development monorepo for Companion for Claude: Obsidian plugin, Claude Code plugin, and guides. Store release: cavi-ai/companion-for-claude.

Install / Use

claude mcp add cavi-ai -- npx -y github:cavi-ai/claude-obsidian

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

79/100

Supported Platforms

Claude Code
Claude Desktop

Our assessment of claude-obsidian

claude-obsidian scores 79/100 on our quality scale, 359th of 549 AI & Machine Learning skills we index.

Its MCP Server is 11 KB long, well organised into 17 sections with 3 code examples: a thorough specification that gives an agent plenty to work with.

It has 3 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
29/30
Structure
18/20
Description
15/15
Adoption
3/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated today, so claude-obsidian is actively maintained.
  • Our last check on 2026-09-02 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 92/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands.

Automated pattern scan on 2026-09-24. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

claude-obsidian compared with similar skills

All 4 of these similar skills score higher than claude-obsidian; compare them before choosing.

SkillScoreStarsUpdatedFormat
claude-obsidian (this skill)by cavi-ai793todayMCP Server
claude-memby thedotmack10094.6ktodayCLAUDE.md
Agent-Reachby Panniantong10085.2k8d agoCLAUDE.md
Understand-Anythingby Egonex-AI10084.0k12d agoCLAUDE.md
headroomby headroomlabs-ai10073.7ktodayCLAUDE.md

Frequently asked questions

How do I install claude-obsidian?
Run claude mcp add cavi-ai -- npx -y github:cavi-ai/claude-obsidian. The install tabs above show the steps for each supported agent.
Which AI agents does claude-obsidian work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is claude-obsidian safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. It is MIT-licensed and scores 92/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is claude-obsidian still maintained?
The repository was last updated today, so claude-obsidian is actively maintained.

claude-obsidian

Cowork with Claude inside your Obsidian vault. Chat with your notes as context, let Claude work the vault with its own tools, render interactive artifacts inline, and drive the same vault from Claude Code.

CI Obsidian downloads License: MIT

▶ Install the Obsidian plugin · Add portable agent workflows · Latest release

Open source · MIT · bring your own Anthropic key · local-first

<!-- hero: assets/hero-daily-rollup.gif — pending capture, see assets/CAPTURE.md -->

| Chat with your vault | Interactive artifacts | |---|---| | Companion chat panel with vault context | A claude-html artifact rendered inline |


What it does

Agent mode

Claude searches, reads, and follows links across the vault while answering, each step shown as a tool chip. Write tools require per-action confirmation. The same agent runs fully local on a tool-capable Ollama model (llama3.1, qwen3) — the composer tells you when the selected model lacks tool support. Optional web search & fetch tools and external MCP servers extend the agent past the vault, both off by default and per-call confirmed. → agent-mode.md

<!-- screenshot: assets/agent-tool-chips.png — pending capture -->

Diff-reviewed edits

Note edits render as a per-hunk red/green diff; only accepted hunks are written. Inline rewrites use the same review path: select text → Rewrite selection with Claude…. Plan Mode restricts a turn to read-only tools and ends in a proposed plan. → agent-mode.md

<!-- screenshot: assets/diff-review.png — pending capture -->

Interactive artifacts

Claude emits a claude-html block; Companion renders it inline in a sandboxed iframe with no vault, network, or cookie access. Open it in a browser or save it as a note. → artifacts.md

A claude-html artifact rendered inline

Research Desk and Workbench

Sources → evidence → claims → outline → draft → assurance, stored as Markdown notes. Only reviewed, locatable, non-stale evidence counts as claim support. Draft revisions are validated before preview and reject unsupported citations. → research-workbench.md

<!-- screenshot: assets/research-desk.png — pending capture -->

Optional MCP bridge

Companion runs a loopback-only, token-gated MCP server over the live vault. The bridge is a specialized integration for explicitly configured MCP clients; it is separate from the CLI-only obsidian-agent plugin. The client direction works too: the agent can consume external MCP servers (HTTP or stdio) from chat, each call confirmed — Companion is the two-way hub. → claude-code-bridge.md

<!-- screenshot: assets/mcp-bridge-settings.png — pending capture -->

Sources, clippings & research import

A watched clippings inbox auto-enriches new clips with typed frontmatter (title, tags, summary — works on a thinking-free local utility call, so bulk ingest stays cheap and reliable) without replacing the clip body or existing metadata. After enrichment, Review all links collects every suggested wikilink into one Git-style review: a collapsed accordion per note, with note-level and per-hunk selection before the accepted set is written. The Organize clippings command takes the existing pile: meaningful renames, tags, and an LLM-inferred domain/project folder per clip, shown as a reviewable move plan before anything is filed. Research sources import from web, PDF, DOI, arXiv, Zotero, or the vault. → research-workbench.md

Local models

The Auto chat backend falls back to a local Ollama model when Claude is offline or out of usage; Local only runs every request on Ollama — including the agent, on models whose metadata reports tool support. Settings badge every detected model with its tools/thinking capabilities, and a composer indicator shows when the current backend reasons before answering. Semantic search uses a built-in on-device embedding model on desktop and mobile — and also indexes vault PDFs, keeping page locators in every result. → local-models.md

<!-- screenshot: assets/local-fallback-indicator.png — pending capture -->

Quick start

  1. InstallSettings → Community plugins → Browse → "Companion for Claude" → Install → Enable, or open it in Obsidian.
  2. Add your key — get one from the Anthropic Console, then paste it in Settings → Companion for Claude → Connection → Anthropic API key and click Save & test connection.
  3. Ask something — open the Companion panel, toggle the Context chip for your active note, and ask a question about it.

Guides


Products and integrations

| Path | What it is | Ships to | |---|---|---| | obsidian-plugin/ | Companion for Claude — the Obsidian community plugin: chat with agent mode, diff-reviewed edits, vault context with PDF/image attachments, link suggestions, consolidated memory, native Canvas/Bases output, inline claude-html artifacts, prompt caching, local-model fallback, and the MCP bridge. | Obsidian community store | | claude-plugin/ | obsidian-agent — universal workflows for Claude, Codex, Gemini, OpenCode, and AgentSkills hosts over the official Obsidian CLI. It has no Companion or MCP dependency. The compatibility-named path is a pinned submodule; see its README. | cavi-ai/plugins and native host packages | | upstream/html-effectiveness/ | Thariq Shihipar's "unreasonable effectiveness of HTML" gallery, vendored as a pinned, unmodified submodule (its own Apache-2.0 license). See NOTICE. | — | | upstream/obsidian-skills/ | Steph Ango's (@kepano) Obsidian Skills, vendored the same way (its own MIT license) and used as the format reference for the Bases, Canvas, and Obsidian-Flavored-Markdown emitters. See NOTICE. | — |

flowchart LR
    companion["Companion for Claude<br/>Obsidian plugin<br/>chat + artifacts · runs the MCP server"]
    bridge(["Loopback MCP bridge<br/>127.0.0.1 · bearer token · port 22360<br/>10 reads · 14 write-gated tools"])
    agent["obsidian-agent<br/>cross-host plugin<br/>official Obsidian CLI"]
    client["Optional MCP clients"]

    companion <-->|"exposes vault tools"| bridge
    client <-->|"specialized live-vault integration"| bridge
    agent -->|"independent portable workflows"| cli["Official Obsidian CLI"]

Ten read/audit tools are always available. Fourteen mutations require Allow writes. Optional web_search and web_fetch join the read set when enabled. In agent mode, writes also keep their per-action confirmation.

Getting it

Install the Obsidian plugin (Companion for Claude)Settings → Community plugins → Browse → search "Companion for Claude" → Install → Enable, or open it in Obsidian.

Install the universal Obsidian agent plugin for Claude Code — enable the official CLI in Obsidian 1.12.7 or newer, then run:

/plugin marketplace add cavi-ai/plugins
/plugin install obsidian-agent@plugins

The universal plugin works without Companion and without MCP. Install Companion separately when you want its in-Obsidian chat, artifacts, local models, or optional live-vault MCP bridge.

<details><summary>Build from source (development)</summary>
git clone --recurse-submodules https://github.com/cavi-ai/claude-obsidian.git
# already cloned without submodules?
git submodule update --init --recursive

See obsidian-plugin/README.md for plugin dev/build steps.

</details>

Open-source hygiene

  • CONTRIBUTING.md — local development, release versioning, security-sensitive review rules.
  • SECURITY.md — supported versions, vulnerability reporting, security boundaries.
  • CODE_OF_CONDUCT.md — contributor standard.

Provenance

The artifact design system is an original reformulation of the aesthetic in Thariq Shihipar's gallery, not a copy of his HTML. The gallery is vendored as a pinned, unmodified submodule.

Steph Ango's (@kepano) Obsidian Skills are vendored the same way at upstream/obsidian-skills/ and serve as the format reference for .base, .canvas, and Obsidian-flavored Markdown generation. His skills operate on vault files; the Companion MCP bridge operates on the running vault. Web-source capture uses his Defuddle library (MIT), the extraction engine behind the Obsidian Web Clipper.

Full attribution is in NOTICE. Everything authored here is MIT-licensed (LICENSE).

License

MIT — see LICENSE. The vendored upstream/html-effectiveness submodule is under its own Apache-2.0 license.

Related Skills

View on GitHub
GitHub Stars3
CategoryAI
Updated12h ago
Forks0

Languages

TypeScript

Trust signals

92/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 low