Obsidian plugins with #mcp#

The Claudian plugin brings AI chat into a vault with support for Claude and other provider setups such as Codex, Grok, Opencode and Pi. It lets you mention notes, folders, MCP servers and external files directly in a conversation, then move between chat tabs, resume older sessions and fork work when a thread needs a new path. Plan mode can inspect a problem before applying changes, while instruction mode adds quick prompt guidance inside the editor. The plugin also keeps shared skills and prompt templates at user and vault scope, stores session metadata locally and exposes settings for provider paths and environment variables.

The Agent Client plugin brings external AI agents into a docked or floating chat inside the vault and keeps each conversation tied to your notes. It can pass wikilink context, pasted files and images to the selected agent, show live terminal output, display note edits as word level diffs and store session history locally so earlier chats can be resumed or forked. The plugin also lets you switch models, modes and settings during a session, approve or reject requested actions, export chats as Markdown notes and run multiple agent sessions in parallel from editor tabs. Notes are read and written through the vault API, while the agent process runs locally with the same system access it would have in your terminal.

The Local REST API plugin exposes a secure local API for working with vault files from scripts, browser extensions and AI clients. It supports full file CRUD, reading or updating the active note, tag queries, command listing and execution, simple text search, and structured search with JsonLogic. A standout part is note patching: you can target headings, block references, or frontmatter keys and append, replace, move, or delete just that section instead of rewriting the whole file. The plugin also ships with an MCP endpoint that mirrors the core API, so external clients can talk to the vault through the same authenticated surface. It includes extension hooks for other plugins to register extra routes and MCP tools.

The MCP Connector plugin exposes your vault to AI clients over the Model Context Protocol through an MCP server that runs inside Obsidian on loopback, with no cloud round trip or separate binary. It includes 52 tools for file reading and writing, search, frontmatter updates, periodic notes, canvas changes, prompts and optional command execution, with structured outputs and tool annotations for safer client behaviour. The plugin also supports per client bearer tokens, per token tool policies and adaptive tool loading so each client can see a narrower or wider tool set without reconnecting. Semantic search runs locally with selectable embedding providers, prompt files can inline note embeds and command execution stays disabled until you allow specific commands per vault.

The Gemini Helper plugin adds a chat interface backed by Google Gemini, along with vault aware tools for reading notes, creating files, proposing edits and searching content. It supports file attachments, slash commands, @ references and model controls for vault access, web search and MCP servers. Beyond chat, it includes a visual workflow builder with node based automation, AI assisted workflow generation, event triggers and execution history. The plugin also includes RAG sync for vault search, OKF knowledge sources, agent skills, external skill packages and plugin based extensions. It tracks token usage, keeps edit history with diff restore, offers optional encryption for chat history and workflow logs and can hand Gemini models and actions to Dashboard Hub and Discussion Hub.

The Local LLM Hub plugin adds a local first AI workspace with chat, workflow automation, RAG search and MCP backed tool use, all routed through locally hosted models or OpenAI compatible endpoints. It can generate or modify workflows and SKILL.md based agent skills from plain language, execute them through a visual node editor, and keep execution history for debugging and retry flows. The chat interface supports attachments, vault tool calls, local MCP servers, reusable skills and semantic search over indexed notes and PDFs. It also integrates with Dashboard Hub and Discussion Hub, tracks AI made edits and can hide encrypted files from chat tools while still allowing workflow access with a password prompt. To use it fully you need a compatible LLM server and an embedding model for RAG.

The Neural Composer plugin adds graph based AI chat to your vault by running a LightRAG server, indexing notes into a knowledge graph and combining graph traversal with vector search for more connected answers. It can start and stop the backend automatically, watch a folder for reindexing, show processing status in the file explorer and open a 2D or 3D graph view for exploration. Answers include citations linked to the source notes and text chunks, and the plugin can expose graph tools through MCP clients. It also supports remote LightRAG servers, multiple model providers and a fully local setup with Ollama, while keeping API keys in the local vault data and avoiding runtime telemetry.

The Large Language Models plugin brings chat based access to cloud and local language models into Obsidian through a sidebar, tab view and floating popup window. It supports hosted providers such as OpenAI, Anthropic, Google and Mistral, while local use can run through GPT4All or Ollama. Models can be switched from each chat surface, and the plugin includes commands to open the different interfaces or toggle the floating action button. It also adds a built in MCP server that can expose vault operations to compatible desktop clients with a local bearer token, confirmation prompts for write actions and localhost only binding. The plugin is aimed at keeping model interaction, local runtimes and vault level tool access inside the same workspace instead of relying on a separate companion bridge.

The ICOR for Life - Connect plugin brings myICOR data into your vault through a sidebar dashboard, knowledge base search, course progress views and room level dashboards computed from local files. It shows journey stages, milestones, reflections, streaks and category rollups, adds AI Team and room dashboards for file activity, and deep links lessons or search results using the Web Viewer setting. It can sync growth assignment reflections into markdown notes as create only files with structured frontmatter, and it registers the myICOR MCP server in `.mcp.json` so Claude sessions in the vault can query lessons, resources and recordings. Authentication uses browser based OAuth with PKCE, tokens stay in the plugin data file and the one time connect flow requires the desktop app.

The Workbuddian plugin connects a local CodeBuddy CLI to your vault and opens an AI chat panel inside the app for streaming conversations, note references and in place writing edits. It supports `@` references for notes, files, subagents and MCP servers from one picker, shows approvals for write, edit, Bash and MCP actions in the chat bubble, and keeps plan mode in the same turn. Vault edits use line level diffs with one click undo and three safety checks. It also supports conversation forking, separate sidebar and main area sessions, visual MCP server management with JSON sync, custom subagents and instant Chinese or English UI switching. It is desktop only, supports Windows and macOS, and requires Obsidian 1.7.2+ plus the WorkBuddy desktop app with the bundled CodeBuddy CLI.