Discover the Best AI Tools

An Obsidian plugin for model chat, note workflows and retrieval using local vector search across several model backends. Model access must be configured, vault tools vary by backend, and local indexing does not prevent cloud models from receiving selected note content.

Lets AI clients search Yandex Lavka products, manage a cart and submit confirmed grocery orders. Uses unofficial private APIs and session cookies, so compatibility may break; real purchases require price confirmation and may need browser payment verification.

Helps Claude search DoorDash menus, compare totals and prepare food orders through chat, a local web interface or MCP. Requires a DoorDash session and compatible runtime; each real order requires user approval.

WeFlow CLI queries and exports authorized local WeChat data and builds reading digests, knowledge search and MCP workflows. AI features use configured local or remote models; local storage does not imply all processing stays offline. Data initialization varies by platform and WeChat version, and remote models can receive selected content under privacy settings.

xlide-mcp exposes Office-file contents, VBA and Power Query for agent inspection and guarded editing, including analysis and macro tests. File-level features can run without Office, but macro execution and certain binary spreadsheet operations require Windows and the desktop application. These actions modify documents or execute code and need deliberate authorization.

Magzdown Agent Tools converts agent-written Markdown into a link displayed as a paginated document by Magzdown. It includes an executable MCP server alongside editor instructions. The text is encoded into the URL rather than uploaded separately; the tool rejects content over 100 KB, and anyone receiving the link receives the content.

Anna is an experimental desktop assistant combining personal tasks, connected business systems and team workspaces with reviewable artifacts. The source preview targets macOS arm64; Jev assignee suggestions require explicit acceptance and do not themselves run the task.

opn-mcp lets assistants shorten and manage links, read analytics and generate QR codes through an opn.onl instance. It requires an API key and a hosted or self-hosted backend; it is a connector, not a standalone shortening service.

Excel Vision MCP reads and edits XLSX/XLSM workbooks, exposing cells, formatting and embedded images to AI clients. It does not calculate newly written formulas; saving may lose pivot tables or complex chart features, and legacy XLS is unsupported.

Markovo converts supported files and explicitly supplied public HTTPS pages into structured Markdown through a CLI or MCP client. Conversion runs on its account-metered service, not locally; credentials and an explicit credit ceiling are required.

TypeSafe Conversation is an unofficial Home Assistant integration that maps Assist commands to built-in device intents using Jev. It sends commands and exposed entities to TypeSafe with your API key; uncertain, conversational or free-text requests can use a configured fallback. It requires Home Assistant 2025.6 or later and currently uses English confirmation templates.

myc provides local task queues and persistent agent memory with an operation log, hybrid search and code indexing. It requires Bun and a supported SQLite version rather than a standard Node.js runtime.

Local-first shared task list and dashboard for multiple coding-agent tools and projects, with claims and optional self-hosted synchronization. Requires Node.js 18+ and an MCP host; no SaaS account is needed.

Local MCP task tracker using expiring claims, resumable checkpoints and version-pinned reviews to coordinate coding agents. Stores each project in SQLite and runs as a Go binary without a cloud account.

OllamaHandoff delegates supplied text to local Ollama models for chat, summaries, code reviews, extraction and commit-message drafts through MCP. It accepts text rather than file paths and requires a running Ollama instance with models installed. Local inference avoids a cloud-model API call, but the calling agent may still incur charges.

logisheets-mcp exposes the LogiSheets calculation engine through a local MCP server. Agents manipulate named blocks, stable columns and row keys, recalculate formulas, preview scenarios and export XLSX workbooks. It requires Node.js 20 or later; Excel-style formulas do not imply complete Excel feature compatibility.

HA-Jev is a Home Assistant integration exposing typed Jev questions through sensors, actions and conversation features, with configurable home checks. It requires Home Assistant 2026.9+ and a TypeSafe API key; relevant home context is sent to the cloud service.

ChatCrystal imports AI coding conversations into structured notes, semantic search, a tag graph and reusable MCP memory. It offers a Windows desktop app and local web/CLI interfaces. Summarization and embeddings require separately configured providers; choosing remote providers sends the relevant content to those services.

LingShu is an alpha-stage desktop agent for code and office documents, with worker sessions, artifact tracking and configurable model providers. It offers macOS and Windows previews; direct computer control is macOS-only and requires authorization. Output quality depends on the model and local toolchain.

NarraNexus is an AI team platform with persistent agent identities, narrative memory, message-based collaboration and per-agent tools. Local and desktop deployments require your own LLM API key; cloud access has a separate trial quota.

Desktop workspace combining agent teams, files, terminal, browser and repository-aware coding. Supports configurable model providers and per-agent permissions. Desktop packages target Windows, macOS and Linux; Computer App Control is limited to Windows and macOS, and WebBridge is a separate companion.

Desktop assistant combining chat, voice, browser and computer tools, MCP connections and background agents. Supports Windows, macOS and Linux. Requires a configured local or hosted model; desktop control depends on OS permissions, and connected cloud services receive the content needed for requests.

MCP adapter for creating, listing, updating and cancelling Pingfyr reminders, including recurring schedules and time zones. Connects to the hosted service for delivery through configured email, webhook, messaging or calendar channels. Requires a Pingfyr API key; local stdio use requires Node.js 18+. The adapter is not a standalone offline reminder scheduler.

MCP server connecting AI clients to Lettio mailboxes over JMAP to read, search, send, reply and organize messages across one or more accounts. The local option requires Node.js 18+ and mailbox app passwords; hosted OAuth access is also documented. Email handling remains subject to the connected AI client’s data practices.