AI plugins for Obsidian 2026: complete comparison (Smart Connections, Copilot, Text Generator, AI Tagger, Companion, CAO)

Choosing among the AI plugins for Obsidian in 2026 means deciding between seven community tools that cater to different needs.

Obsidian still doesn’t include any native artificial intelligence: everything is through community plugins, and the decision matrix is structured around four intersecting axes.

Task type, support for local LLMs, economic model, and vault integration differentiate a free offline Smart Connections from a Copilot Plus at $14.99/month and a Smart Composer still in BRAT beta.

This comparison examines the seven major plugins, suggests a combo for each profile (writer, researcher, developer, strict GDPR compliance), and shows how popular non-AI marketplace components become a foundation for AI workflows.

Why Obsidian still lacks native AI in 2026

Obsidian remains a local-first Markdown editor, with each note existing as a .md file on the user’s disk.

The Dynalist team maintains the file-over-app philosophy: no proprietary cloud, no lock-in, no enforced SaaS.

The engine remains minimalist, with intelligence coming from the marketplace, which exceeds 2,000 community plugins, without LLM integration on the editor side.

Three AI families structure this landscape: semantic RAG search, assisted text generation, and automated tagging.

The first family uses embeddings (semantic vectors) to link notes with similar meanings.

The second family uses a chat LLM (cloud or local) to produce text on command.

The third family combines both: extracting entities, assigning tags, feeding the graph.

The analogy is simple: Obsidian acts as a personal knowledge OS, with AI plugins adding specific capabilities without altering the core.

The absence of native AI is not a delay, it’s a design choice: as long as the vault remains 100% Markdown readable without software, any plugin can index, generate, or tag without breaking portability.

The seven major AI plugins for Obsidian across four axes

The chosen matrix crosses task type, local LLM support, economic model, and vault integration.

The table below summarizes the seven reference AI plugins for Obsidian, with author or repository, main type, RAG vault, local model support (Ollama, LM Studio), and cost.

Smart Connections: free local embeddings

Brian Petro has positioned Smart Connections as the vault’s private semantic search engine.

The plugin includes a local embedding model that automatically indexes each note without an API key.

The sidebar surfaces notes with similar meanings during typing, and a chat view allows questions to be asked to the vault with source citations.

The immediate analogy: a Spotlight bar enriched with meaning, working without an internet connection.

Copilot for Obsidian: freemium and Vault QA

Logan Yang offers a Free tier (BYOK, no registration) and a Plus tier at $14.99/month ($11.67/month annually).

The Vault QA mode relies on a local vector index and accepts Ollama or LM Studio to remain 100% offline.

The Plus plan unlocks integrated chat-model, web and YouTube search in chat, and an autonomous agent with long memory.

The Self-Host Supporter option at $349.99 lifetime includes two years of Plus and then switches to self-host with your own keys.

Text Generator: the AI printer cartridge

Nour Aouari delivers a fully open-source, multi-provider plugin (OpenAI, Anthropic, Gemini, Ollama, OpenRouter).

The template engine stores each prompt as a reusable Markdown file from any note.

The analogy is a Word macro boosted by LLM, where the cartridge changes depending on the connected model.

AI Tagger Universe: the automatic librarian

Nie Hu targets large-scale tagging: the plugin reads a note’s content and suggests predicted tags.

The tags then feed into the Obsidian graph, enhancing theme-based navigation.

The role remains assistant: the user validates or rejects, never mass writing without confirmation.

Companion: ghost text autocompletion

rizerphe has brought the GitHub Copilot pattern to Obsidian, but for general text, not code.

During typing, the plugin suggests the likely continuation of the sentence in light gray, confirmable with the Tab key.

The rendering remains discreet, without intrusive context menus, which suits long writing well.

Smart Composer (CAO): Cursor for notes

Smart Composer replicates the Cursor-like editing experience in the vault, with context drawn from neighboring notes and one-click edits.

The plugin remains in BRAT beta at the time of writing: to be installed manually via the Beta Reviewers Auto-update Tester plugin.

The approach mirrors the Cursor approach for developers but applied to Markdown notes instead of source code.

Flashcard Generator: the Anki card factory

ChloeDia has released a plugin that transforms a note into a series of question-answer pairs ready for Spaced Repetition.

Export to Anki or the Obsidian Spaced Repetition plugin is done in one click.

The format remains readable: each card exists as a Markdown block, versionable and auditable like the rest of the vault.

Which AI plugin combo for your profile

The useful recommendation doesn’t point to a winning plugin, it assembles a coherent trio around a dominant use.

Writer or content creator: Smart Connections to link notes by meaning, Text Generator connected to Templater for structured note generation, Companion for fast typing in autocomplete mode.

The trio covers material discovery, text production, and writing speed without heavy cloud dependency.

Researcher or student: Smart Connections for mapping literature, Copilot for querying the vault with cited sources, Flashcard Generator for reviewing via Spaced Repetition.

This combo creates a reading, restitution, memorization loop without leaving Obsidian.

Developer or technical editor: Smart Composer (CAO) for selection-based rewrites, Copilot for assisted Dataview queries, Text Generator for README and changelogs.

Smart Composer brings the visual diff experience closer to an IDE.

Strict GDPR profile (lawyer, medical, legal, regulated teams): Smart Connections with local embeddings, Copilot connected to Ollama or LM Studio, sensitive folders excluded from the index.

The result fits on an M-series MacBook: a personal sovereign data center without sending to a third-party cloud, scenario detailed for building an Obsidian knowledge base with Claude.

The right combo isn’t chosen by adding features, it’s chosen by removing any plugin that doesn’t serve the profile’s main movement: a writer doesn’t need a Vault QA, a researcher doesn’t need ghost text.

Popular non-AI components as the foundation of AI workflows

Excalidraw, Templater, Dataview, Tasks, Advanced Tables, Calendar, Git, Kanban, Style Settings, and Iconize remain non-AI by nature, which is precisely what makes them solid foundations.

AI is grafted on top via documented combos, never the reverse.

Excalidraw plus ExcaliAI 2.0: from sketch to clean diagram

Zsolt Viczian maintains Excalidraw, one of the most downloaded on the marketplace.

ExcaliAI 2.0 adds two key functions: text-to-diagram (description in prompt, clean diagram output) and wireframe-to-code (sketch to HTML).

The analogy is simple: a whiteboard that transforms into a clean diagram via prompt.

Templater plus Text Generator: the production line

Templater manages dynamic scripts in Obsidian templates, without LLM calls by default.

Connected to Text Generator, each template opens a model call at fill time: auto-filled reading note, auto-generated metadata, footer summary.

Dataview, another major component, provides SQL-like queries on the vault, and Copilot or Smart Composer helps write them without full syntax knowledge.

Tasks, Git, Kanban, and transversal components

Tasks structure actions in notes, and an AI agent can extract these tasks from a meeting transcript and prioritize them.

Obsidian Git versions the vault like a Git repository, which becomes a mandatory discipline before letting an agent rewrite dozens of notes.

Advanced Tables, Calendar, and Kanban manage everyday data structures (tables, dates, columns), and AI connects to summarize, prioritize, or reclassify.

Style Settings and Iconize aren’t AI tools, but visually marking the AI zones of the vault (icons, colors) helps trace what AI touches.

Privacy, monthly cost, and choosing a local LLM

The cost question breaks down into two independent layers: embeddings and chat.

Smart Connections runs its embeddings 100% locally by default: zero API cost even on a vault with 10,000 notes.

Copilot and Smart Composer open the chat layer: with an OpenAI or Anthropic key, the average usage ticket remains around a few USD per month for moderate use (source Effortless Academic, order of magnitude of $0.01 for 1,000 GPT-4o tokens mid-2025, to be confirmed in 2026).

The 100% local mode eliminates the cloud bill: Ollama or LM Studio serve the model, Smart Connections serve the embeddings, the pipeline fits on an M-series MacBook or a PC with a suitable GPU card.

The guide to installing a local model with LM Studio covers the step-by-step setup for DeepSeek-R1, transferable to Qwen, Llama, or Mistral.

On mobile, local indexing remains heavy: issue Copilot #165 documents iOS and Android limits, and the practice is to index on desktop then sync the vault.

For strict GDPR, the complete chain holds on three components: Ollama for chat, Smart Connections for embeddings, sensitive folders excluded from the index via plugin configuration.

Versioning before launching an AI agent on the vault isn’t a precaution, it’s a professional discipline: an activated Obsidian Git saves reversibility, without which a script rewriting 300 notes becomes a net loss.

Conclusion

The landscape of AI plugins for Obsidian in 2026 consists of seven active components, two independent layers (embeddings and chat), and a structural choice between cloud BYOK, freemium Copilot, and 100% local chain.

The four-axis matrix (task, local LLM, economy, vault integration) provides the filter, the user profile provides the prescription.

Anthem is preparing six satellite articles to delve into each component: Smart Connections tutorial, step-by-step Vault QA Copilot, Templater plus Text Generator, Ollama configuration for Obsidian, researcher workflow with cited sources, student workflow with Spaced Repetition.

The hub is completed as publications are added in the Artificial Intelligence section of anthemcreation.com.

FAQ

What is the best free AI plugin to start in Obsidian without an API key?

Smart Connections: it includes a local embedding model by default, automatically indexes the vault, and displays notes with similar meanings without any API key.

Smart Connections or Copilot: which to choose for interacting with my notes?

Copilot better covers the chat-vault scenario with its Vault QA mode and autonomous agent (Plus), while Smart Connections excels in pure semantic search with quick source citation.

Both are often used together: Smart Connections for discovery, Copilot for conversation.

How to run a 100% local LLM in Obsidian (Ollama, LM Studio)?

Install Ollama or LM Studio on the machine, download a model suited to the available RAM (Llama 3.1 8B, Qwen 2.5 7B, DeepSeek-R1 8B), then in Copilot or Smart Composer configure a custom local endpoint pointing to the server API (usually localhost:11434 for Ollama).

What is the real monthly cost of intensive AI use on a large vault?

A vault with several thousand notes and daily chat use runs around a few USD to a few tens of USD per month in BYOK depending on the model used (GPT-4o at $0.01 / 1,000 tokens mid-2025, to be confirmed in 2026).

The local mode eliminates this recurring cost, at the price of an initial hardware investment.

Which AI plugins work on Obsidian mobile (iOS, Android)?

Smart Connections supports mobile but local indexing remains heavy on large vaults.

Copilot issue #165 documents the limits on Copilot’s side, and the recommended practice is to index on desktop then sync the vault via Obsidian Sync, iCloud, or Git.

How to protect the privacy of my notes when connecting a cloud LLM?

Three levers: exclude sensitive folders from the plugin index, switch to local mode (Ollama plus Smart Connections local embeddings) for critical areas, and check the provider’s retention policy (Brevilabs on Copilot Plus announces zero content retention).

Which AI plugin combo for a researcher, student, developer, or writer?

Writer: Smart Connections plus Text Generator plus Companion.

Researcher or student: Smart Connections plus Copilot plus Flashcard Generator.

Developer: Smart Composer plus Copilot plus Text Generator.

Can you generate Anki flashcards directly from your Obsidian notes via AI?

Yes: Flashcard Generator (ChloeDia) reads the active note, suggests question-answer pairs generated by an LLM, then exports to Anki or the Obsidian Spaced Repetition plugin.

How to combine Templater and Text Generator to automate reading notes?

Define a Templater template with placeholders (title, author, summary), connect Text Generator to a Templater command that calls a prompt stored in a note, and link the trigger to a keyboard shortcut.

The note fills in one keystroke from a pasted source text.

What risk do you run by letting an AI agent modify hundreds of notes without Git?

Net data loss: an agent rewriting or deleting passages without versioning leaves no reliable rollback.

The professional discipline is to activate Obsidian Git, commit before any execution, and isolate agent operations in a dedicated branch if usage is intensive.