That distinction matters. Saying "Obsidian has no AI" is now too broad. Saying "Obsidian has built-in AI" suggests a vault-wide assistant that the desktop app does not provide. The useful answer depends on which job you want AI to do.
What AI does Obsidian include today?
The current official core-plugin list includes tools such as Audio recorder, Bases, Canvas, Graph view, Search, and Templates. It does not list a general AI chat, generative writing, vault question-answering, or AI meeting-summary feature.
Obsidian's official Web Clipper Interpreter is the important exception. Interpreter is a feature of the Obsidian Web Clipper browser extension. It can use natural-language prompts to extract, summarize, explain, translate, or reformat a web page before saving the result to Obsidian.
Interpreter is not an Obsidian-hosted model and not a vault-wide assistant. You choose a language-model provider or a local model, configure it, and run the prompt while clipping a page. Obsidian says requests go directly to the selected provider and that Obsidian does not gather or store the requests. A local model served through Ollama can keep that processing on the device.
| Question | Answer |
|---|---|
| Does the Obsidian desktop app have general AI chat or writing built in? | No, not in the current core-plugin set. |
| Does Obsidian offer an official AI-powered feature? | Yes. Web Clipper Interpreter can process web pages with a model you configure. |
| Can AI search, summarize, or write across a vault? | Yes, through community plugins or external tools, with capabilities and data handling that vary. |
| Can Obsidian create AI meeting notes on its own? | The core Audio recorder saves audio, but core Obsidian does not provide AI meeting transcription or summaries. |
Three ways to use AI with Obsidian
1. Use Web Clipper Interpreter for web research
Choose Interpreter when the input is a web page and the output should become a structured note. It can extract a paper's title and claims into properties, summarize a long reference page, translate a passage, or make repeated research captures follow one template.
The boundary is clear: Interpreter processes the page you are clipping. It does not automatically read the rest of your vault, join meetings, or act as a general assistant inside the desktop app.
Review the model configuration before using it with private or licensed material. A cloud provider receives the page context sent in the request. A local model changes that boundary, but adds setup, model-quality, and hardware tradeoffs.
2. Add a community plugin for AI inside the vault
Obsidian's community-plugin system lets third-party developers add features and integrate third-party services. Specific AI capabilities and data paths depend on the plugin.
This is the route to consider when AI needs to work from notes already in the vault. It also requires careful evaluation. Obsidian warns that community plugins run third-party code on your behalf, and Restricted mode must be turned off before you can install them.
Obsidian cannot reliably limit community plugins to granular permissions: they inherit the app's access and may access files on your computer, connect to the internet, or install programs. Community plugins also do not update automatically, so check for and install updates manually.
Before enabling an AI plugin, check five things:
1. Scope: Which folders or files can it read and modify? 2. Provider: Does it call a cloud API, a local model, or both? 3. Index: Where are embeddings or other search indexes stored? 4. Evidence: Does an answer link back to the notes it used? 5. Maintenance: Is the plugin actively updated, and can you recover if it stops working?
For specific options, use the AI plugins for Obsidian guide. If your goal is asking questions across existing notes, the chat with your Obsidian vault guide explains the retrieval layer separately.
3. Use an external tool to create better source notes
The third path leaves the AI outside Obsidian. An external tool captures or processes work, writes a normal Markdown file, and lets Obsidian store and connect the result.
This is often the cleaner architecture for meetings. A plugin running inside Obsidian is not automatically the best place to capture system audio from another call app. The useful boundary is the file: the capture tool creates a reviewable artifact, and Obsidian remains the system of record.
Shadow is one example. Shadow is an AI interface for Apple Silicon Mac that sees, hears, and runs. It is a separate Mac app, not an Obsidian plugin. Core meeting capture, transcription, diarization, and the local meeting record stay on the Mac. Shadow saves Markdown and available media in its vault folder, which Obsidian can open directly as a vault.
When a Meeting Skill runs, relevant meeting context is sent for external AI processing, and a supported Skill can save a Markdown copy of the result in the meeting folder. That is different from local capture alone. Review what stays local and what can leave your Mac before enabling automated Skills or webhooks.
For a current capture-to-vault comparison, see the AI meeting assistant for Obsidian guide.
Which setup should you choose?
Start with the source material, not the model brand.
| You want to... | Start with... | Main boundary to verify |
|---|---|---|
| summarize or structure a web page before saving it | Obsidian Web Clipper Interpreter | selected model provider and page context |
| chat with or rewrite notes already in the vault | a reviewed community plugin | file access, indexing, provider, and citations |
| keep cloud AI out of a sensitive vault | a vetted plugin configured for local indexing and a local model | storage, network paths, device resources, and plugin code |
| turn meetings into durable Markdown | an external capture workflow such as Shadow | capture permissions, consent, AI-processing settings, and output folder |
| use Obsidian with no generative AI | core Obsidian only | services and plugins you separately enable |
This modularity is one of Obsidian's strengths. You do not have to grant one assistant access to everything. Web research, vault search, meeting capture, and writing can have different tools and different privacy rules.
A safer setup sequence
1. Decide what the authoritative record is
Obsidian stores notes as Markdown-formatted plain text files in a local vault. Keep source notes readable without an AI plugin. Generated summaries should point back to transcripts, research clips, or original notes rather than replace them silently.
2. Add one AI boundary at a time
Start with one narrow job, such as summarizing web pages or searching a non-sensitive folder. Test it before enabling access to an entire work vault. Several overlapping AI plugins make it difficult to know which tool changed a file or sent context externally.
3. Test with representative, non-sensitive material
Use a real note shape, not a perfect demo document. Check whether the tool preserves links, properties, headings, and source references. For chat, verify material answers against cited notes. For generation, review the diff before keeping the change.
4. Document the data path
Write down what remains local, what is indexed, what is sent to a provider, and what is stored after the request. "Local vault" does not make every installed plugin or connected model local.
5. Keep a recovery path
Back up the vault before letting any plugin or external tool edit files in bulk. Obsidian's File Recovery keeps limited, device-local snapshots, and sync is not a backup. Keep a separate current backup or reliable version history before allowing bulk edits.
How Shadow and Obsidian fit together
Use Shadow when the missing input is what happened in a meeting, not when the task is simply chatting with notes that already exist.
1. Install Shadow on a supported Apple Silicon Mac and complete the required Microphone, System Audio, Screen Recording, and Accessibility permissions. 2. Choose a dedicated Shadow vault folder. A folder used only by Shadow is the simplest setup, and Obsidian can open it as a vault. 3. Start Listening manually, use the quick start and stop shortcut, or enable Autopilot for supported meetings. 4. Review the saved transcript, notes, audio, screenshots, and meeting metadata available for that meeting. 5. Enable a Meeting Skill only when its external-processing behavior and output destination are appropriate. Review generated decisions, owners, dates, and claims before sharing them.
The division of labor is simple: Shadow captures and creates the meeting record; Obsidian stores, links, and surfaces the files. An optional AI plugin can then search or synthesize across those files.
FAQ
Does Obsidian use AI automatically?
No. The desktop app does not automatically run a general AI assistant over your vault. Web Clipper Interpreter is optional and runs only when configured and invoked. Community plugins and external tools are also separate choices that you enable.
Is Obsidian Web Clipper Interpreter built-in AI?
It is an official AI-powered feature of the separate Obsidian Web Clipper extension. It is not a general assistant inside the Obsidian desktop app. You configure the model connection, and the request can go to a cloud provider or a local model.
Does Obsidian have AI meeting notes?
Core Obsidian includes an Audio recorder that saves audio in a note, but its current core-plugin set does not provide AI meeting transcription or summaries. A community plugin or external meeting-capture tool can add that workflow.
Can I use AI with Obsidian without sending my vault to a cloud model?
Potentially. Web Clipper Interpreter officially supports a local model. For vault-wide AI, use a plugin only after verifying that both indexing and generation are local and that no other network calls transmit note content. A plugin may index locally but still send retrieved text to a cloud model.
Is Shadow an Obsidian plugin?
No. Shadow is a separate Mac app. It can save its local meeting record and supported Markdown outputs in a folder that Obsidian can read. That file-based handoff keeps Obsidian as the place where notes are stored and connected.
Sources
- Obsidian core plugins, for the current built-in desktop feature list.
- Obsidian Web Clipper Interpreter, for prompt behavior, provider types, local-model options, and request handling.
- Obsidian community plugins, for installation, Restricted mode, updates, and third-party-code warnings.
- How Obsidian stores data, for local vault and Markdown-file behavior.
- Shadow vault and local files, Meeting Skills, and privacy and data leaving your Mac, for Shadow's current file, Skill-output, and processing boundaries.
This article was written by Chad Oh, Shadow's AI writer. While we strive for accuracy, AI-generated content may contain errors. If you spot something off, let us know.