Obsidian AI Plugins: Copilot vs Smart Connections vs Others [Radar]

Copilot for Obsidian is the best all-round AI plugin; Smart Connections is the best for passive recall. Across six criteria — capability, retrieval quality, local-model support, setup effort, cost, and ecosystem maturity — Copilot scores highest for users who want one tool that chats with their vault and generates text, while Smart Connections wins for surfacing related notes automatically with no prompting. Text Generator is the pick for templated, prompt-driven drafting. The single dimension that should drive your choice is whether the plugin can run on a local model, because that decides whether your notes stay on your machine.

A note on how this comparison works. We The Flywheel is vendor-agnostic. None of these plugins paid for placement or influenced its score — they are free, open-source community plugins, and we assessed each one against the same rubric. The scores below come from independent analysis as a working knowledge base, not from project READMEs. We run our own evaluations and score against a fixed rubric. The full framework is documented and public, and the scoring algorithm is detailed in How We Score Tools.

How do the Obsidian AI plugins score head-to-head?

We scored five widely used plugins across six dimensions that define an AI layer on a note vault: what it can do, how well it retrieves the right notes, whether it can run against a local model, how much setup it demands, what it costs to run, and how mature and maintained it is. Each dimension is scored out of 10. The composite is the unweighted mean, because the right weighting depends on your job — a privacy-driven user should weight local-model support heavily; someone who just wants ambient links should weight setup effort and retrieval.

Two plugins separate from the field. Copilot for Obsidian takes the top composite because it covers the most jobs competently in one install. Smart Connections sits a fraction behind despite a narrower scope, because it does its one job — ambient recall — exceptionally and installs with almost no friction. The rest are specialists: reach for them when their single strength matches your primary need, not as a default. Read the per-plugin sections before trusting the headline number.

What can each plugin actually do?

Capability is where the field splits into three jobs: chat, recall, and generation. Naming the job first is the fastest way to a shortlist.

Copilot for Obsidian is the broadest. It opens a chat panel that answers questions over your vault using retrieval-augmented generation, runs semantic search, and generates or rewrites text inline. It supports a range of cloud providers and local models, which is why it earns the top capability score — it is the closest thing Obsidian has to a built-in assistant. Smart Connections is deliberately narrower: it embeds your notes and continuously surfaces semantically related ones in a sidebar as you write, plus a chat feature, but its core value is passive discovery rather than generation. Text Generator inverts the model — it is built around templates and prompts, so you trigger generation from a structured prompt rather than a conversation, which suits repeatable drafting workflows. Smart Second Brain and BMO Chatbot are lighter chat-focused plugins that lean local-first. Pick the capability profile that matches your job before comparing anything else.

Which retrieves the right notes best?

Retrieval quality is the dimension users feel most and discuss least. An AI plugin that generates fluent text from the wrong notes is worse than useless — it is confidently wrong about your own knowledge. The differentiator is how each plugin embeds and chunks your vault and how it ranks what to feed the model.

Smart Connections and Copilot lead here because both build a proper embedding index over the vault and do semantic similarity ranking, so they surface notes that are conceptually related even when they share no keywords — the entire point of a second brain. Text Generator and the lighter chat plugins lean more on the active note or a manual selection as context, which is fine for drafting but weaker for "find what I wrote about this across 4,000 notes." If your vault is large and your goal is recall across it, weight retrieval heavily and the shortlist narrows to two. Confirm current embedding models and chunking behavior on each plugin's repository, because retrieval internals change frequently.

Can it run locally, and why that should drive your choice

This is the dimension that should decide a privacy-sensitive choice, and the one most buyers underweight. Every plugin here can point at a cloud API — OpenAI or Anthropic — which gives you frontier model quality but sends your note content to a third party. Not every plugin runs fully against a local LLM.

Copilot for Obsidian, Smart Connections, Smart Second Brain, and BMO Chatbot all support local models through Ollama or LM Studio, so embeddings and chat run on your own machine and notes never leave it. Smart Second Brain in particular is built local-first, which is why it scores highest on this dimension despite a narrower feature set. The trade-off is real: a local 7B–14B model is meaningfully weaker than a frontier cloud model and slower without a capable GPU. For a vault holding client work, regulated data, or anything you would not paste into a public tool, that quality hit is the correct price for keeping your knowledge base on your own hardware. Choose privacy first; choose features second.

What does setup and running cost look like?

The plugins are free and open-source; the cost is the model behind them. Point a plugin at a cloud API and you pay that provider's per-token rate, which for moderate personal use typically lands in the low single-digit dollars a month. Run a local model through Ollama and the model is free — you pay only in hardware and electricity. There is no plugin license to budget for.

Setup effort is where the field separates again. Smart Connections is close to install-and-go: it works with minimal configuration, which is why it scores highest on ease. Copilot and Text Generator reward tuning — models, prompts, retrieval settings — so they take longer to reach their full value but go further once configured. The local-first plugins add the one-time step of installing Ollama or LM Studio and pulling a model. None of this requires writing code; the heaviest technical task is pasting an API key or running an installer. If you want a fully bespoke retrieval layer instead of a plugin, that is a build project — see the companion build below.

The verdict, by job-to-be-done

There is no universal winner. There is a right plugin for each job. Match yours below.

Pick Copilot for Obsidian if…

• You want one plugin that does most jobs — chat with your vault, semantic search, and inline generation in a single install.
• You will tune it. It rewards configuration of models, prompts, and retrieval, and goes furthest once set up.
• You want model flexibility — switch between cloud providers and a local model without changing plugins.
• Chat-with-your-notes is the core workflow, not an occasional convenience.

Pick Smart Connections if…

• You want recall, not generation — notes that surface themselves as you write, with no prompting.
• You value a near-zero setup and a lightweight plugin that works almost out of the box.
• Your vault is large and rediscovering your own past thinking is the primary job.
• You will pair it with an active plugin for the times you do want chat or generation.

Pick a specialist if…

• Text Generator — your workflow is templated, repeatable drafting from structured prompts rather than open conversation.
• Smart Second Brain — privacy is paramount and you want a chat-with-vault experience that runs local-first by design.
• BMO Chatbot — you want a lightweight in-editor chat with local-model support and minimal overhead.

Or run two

The most common setup among practitioners is not one plugin but two: Smart Connections for ambient, zero-prompt recall, plus Copilot or Text Generator for on-demand chat and generation. They occupy different jobs and do not conflict, and the pairing beats forcing a single plugin to cover both passive discovery and active generation. If you would rather build your own retrieval layer — your own embeddings, chunking, and Ask-me interface over your notes — see our companion build on a local RAG over an Obsidian vault in Python, and the broader second-brain pillar for the system view. For the tool-choice layer underneath all of this, see Obsidian vs Notion.

The bottom line

The Obsidian AI plugin field is not a single race — it is three jobs (chat, recall, generation) with different winners. Copilot for Obsidian takes the top composite because it covers the most jobs in one install; Smart Connections sits just behind because it does its one job, ambient recall, better than anything else and installs with almost no friction. The specialists win only when their single strength is your primary need.

Decide the job before the plugin, and let local-model support break ties when privacy matters. If your vault holds anything sensitive, weight that dimension above feature breadth and the shortlist resolves itself. We update these scores as the plugins ship releases — community plugins move fast. Check the Technology Radar for the latest, read the Obsidian vs Notion comparison for the tool layer beneath the plugins, and see the full methodology for how we arrive at these recommendations.

What is the best AI plugin for Obsidian in 2026?

There is no single best plugin — the right one depends on the job. For an all-in-one assistant that chats with your vault, runs semantic search, and generates text inline, Copilot for Obsidian scores highest on our rubric and supports both cloud and local models. For passive recall — surfacing related notes automatically while you write, with no prompting — Smart Connections is the lighter, higher-value pick. For templated, prompt-driven generation (expanding outlines, drafting from a structured prompt), Text Generator is purpose-built. Most serious users run two: Smart Connections for ambient linking plus Copilot or Text Generator for active generation. Match the plugin to whether your primary need is recall, chat, or generation rather than expecting one to do all three well.

Can Obsidian AI plugins run on a local LLM without sending notes to the cloud?

Yes, several can, and this is the most important dimension for a private vault. Copilot for Obsidian and Smart Connections both support local models through Ollama or LM Studio, which means embeddings and chat run on your own machine and note content never touches a third-party server. Text Generator also supports local and self-hosted endpoints. The trade-off is quality and speed: a local 7B–14B model is noticeably weaker than a frontier cloud model and slower without a capable GPU. If your vault holds client work, research under NDA, or anything you would not paste into a public tool, the local-model path is worth the quality hit. Verify each plugin's current local-model support on its repository before committing, because this capability changes release to release.

What is the difference between Smart Connections and Copilot for Obsidian?

They solve different problems. Smart Connections is a passive recall engine: it embeds your notes and continuously surfaces semantically related notes in a sidebar as you write, with no query required — its job is to help you rediscover your own thinking. Copilot for Obsidian is an active assistant: you open a chat panel, ask questions of your vault (retrieval-augmented), and get generated answers, summaries, or drafts on demand. Smart Connections is lighter and ambient; Copilot is heavier and conversational. They are complements, not substitutes — a common setup runs Smart Connections for ambient linking and Copilot for on-demand chat and generation. If you only install one, choose by whether you want notes to find you (Smart Connections) or to interrogate your notes (Copilot).

Are Obsidian AI plugins free?

The plugins themselves are free and open-source — Copilot for Obsidian, Smart Connections, and Text Generator are all community plugins installed at no cost. What you pay for is the model. If you point a plugin at OpenAI or Anthropic, you pay that provider's per-token API rate, which for moderate personal use is typically a few dollars a month. If you run a local model through Ollama, the model is free and you pay only in hardware and electricity. Some plugins also offer optional paid tiers or hosted convenience layers (for example, a managed embedding service), but the core functionality runs on your own API key or local model. Budget for API usage, not for the plugin. Confirm any optional paid tiers on each plugin's page, as monetization models for community plugins have shifted over the past year.

Do I need to know how to code to use an Obsidian AI plugin?

No coding is required to install or use these plugins — they install from Obsidian's community plugin browser in a few clicks and are configured through settings panels. The one technical step most plugins ask for is pasting an API key from OpenAI or Anthropic, or installing Ollama for local models, both of which are guided processes rather than programming. The configuration depth is where effort scales: Smart Connections works almost out of the box, while Copilot and Text Generator reward tuning prompts, models, and retrieval settings. If you want a fully bespoke retrieval layer over your notes — your own embeddings, your own chunking, your own Ask-me interface — that is a build project rather than a plugin install; see our companion piece on building a local RAG over your notes for that path.