Discover the Best AI Tools

Kiln gives AI agents tools to design parts, slice models, queue and monitor jobs on supported 3D printers. It needs compatible printer and design software setup and human acceptance of terms; advanced services and external manufacturing can incur charges.

KiCad MCP Server lets AI assistants edit schematics and PCB layouts, inspect libraries, run electrical and design checks, and export manufacturing files. It requires KiCad and related toolchains; automated checks do not replace engineering review.

Memi analyzes frontend repositories to provide file-linked design and accessibility findings and a focused interface brief for coding agents. Local diagnosis uses deterministic rules rather than model calls. Its newer frontend workflow is beta and Canvas remains in development; source checks are not complete visual certification or proof of cost savings.

localfig connects an MCP client to Figma Desktop through a local bridge and development plugin. It can inspect and modify frames, text, layouts, variables and components using the Plugin API. The desktop plugin must remain running, and edits affect the active file; the bridge does not make Figma or the AI provider entirely offline.

MCP design engine integration for editable layouts, asset imports and print-ready exports. Requires Node.js 22.15+; the repository is MIT-licensed but the engine is proprietary, Windows needs WSL, and image-to-layer conversion requires IMG.LY sign-in.

Lets assistants list, read, create and update Mermaid diagrams in a Diagramzu Space, inspect structural summaries and retrieve versions. Requires a Diagramzu account and a token scoped to one Space. Diagrams are stored in the hosted service; local stdio is a connection option, not offline storage.

Models journeys as state machines and screens as contracts, checking reachability, missing loading/error states, contrast, touch targets and text expansion. Includes previews, exports and implementation-drift audits for CI. Checks depend on the declared design and supported rules; synthetic benchmark accuracy does not establish complete accessibility or real-world usability.

Provides deterministic color conversion, mixing, harmonies, perceptual differences, text-contrast checks and saved palettes through MCP. The public endpoint currently disables authentication; self-hosted deployments can enforce OAuth scopes for reading and palette writes. Contrast checks cover color ratios, not a complete accessibility audit.

A thin stdio client for the Diagrams.so API, enabling assistants to generate, edit and manage cloud architecture diagrams, inspect warnings and export draw.io files. Requires Node.js 18 or later and browser-approved account access. It uses the hosted production API by default; logging in alone does not generate a diagram.

A visual document workspace with HTML/CSS previews, annotations, brand assets and schema-backed collections for repeated documents, plus PDF export. Documents retain editable state and versions. Optional Gmail integration requires the user’s OAuth setup and creates drafts only; it does not send email.

A browser canvas and MCP engine for calibrated measurements, material takeoffs, markups and exports from construction PDFs, retaining geometry provenance and review status. Sheets require calibration and quantities need review. One-click room detection is currently gated and unavailable by default; the open engine does not include the proprietary training data or models.

Searches and ranks design kits, explores names and domains, and exports CSS, Tailwind, shadcn or DTCG tokens. Can apply design files with hash-based conflict checks. Public kits are accessible without an account; persistent work and Pro capabilities require authentication or a paid tier. Fit scores do not guarantee design suitability, and domain availability is not trademark clearance.

Generates and iterates iOS-style interface mockups as PNG and HTML/Tailwind using Claude and Playwright Chromium, with a local design library. Requires an authenticated Claude CLI subscription or an Anthropic API key. Outputs are interface prototypes, not native iOS applications; the local CLI still uses a remote model.

A local MCP design toolkit combining reference documents, token and color generators, design-system workflows and design/copy linters. Its source checks cover only part of the reference library; curated guidance is not uniformly verified. Local tool execution does not determine whether the connected AI client sends data externally.

Builds, validates and repairs workflow, sequence, architecture and mind-map diagrams using FlowZap’s own DSL and public APIs. It does not use Mermaid syntax or access private account diagrams. Remote creation and optional LLM compliance checks are rate-limited; those checks are suggestions, not compliance certification.

Serves a user-authored brand system to AI clients, including verbal guidance, colors, typography, tokens and components. Requires populated brand files and a compatible v2 configuration. Most tools are read-only, but a local document-sync tool writes files; it does not invent a complete brand from an empty template.

Creates physical, VLAN and IP network diagrams and device tables from structured design data, with local MCP, online and offline editions. Local MCP exports SVG, PowerPoint and HTML while keeping master files locally. Live-inventory conversion uses a separate extension; diagrams do not themselves configure or validate a production network.

A bidirectional MCP server connecting coding agents to a Figma plugin through a local relay, exposing design context and canvas edits. Requires Figma, the plugin and compatible Node.js. The relay is local, but design information passed to a connected cloud AI follows that client’s data handling; it is not an end-to-end offline guarantee.

An MCP client for PicoBerry’s hosted image and 3D asset pipeline, including remeshing, texturing and animation. Requires API access enabled by a completed purchase and generation credits. Assets are drafts for review and refinement; engine support, export options and costs vary by operation.

Queries, aggregates and inspects live Autodesk Revit models through a workstation add-in and MCP client. Read-only by default; MCP writes require both server and workstation opt-in gates. Requires a Windows Revit workstation, and requested model data is returned to the chosen AI client.

Controls 2D drafting and 3D solids through an AutoLISP file bridge, including geometric measurements and batch undo. The reference setup requires Linux, BricsCAD and xdotool; other CAD adapters are prospective rather than verified equivalents. Arbitrary LISP execution can change drawings, and measurements are not engineering certification.

A CAD toolbox for MCP assistants to create build123d models, render previews, measure geometry and export STEP, STL, SVG or DXF. Requires compatible Python/CAD dependencies and a separate assistant. It is not a standalone chatbot, and reported benchmark gains do not guarantee correctness of every generated part.

Connects an AI assistant to TouchDesigner to create, wire, inspect and preview node networks using an embedded operator reference. Requires TouchDesigner and an in-app bridge. The bridge executes project operations and listens on all network interfaces by default, so its access must be restricted to trusted networks.

A CLI and MCP interface driving FreeCAD Python workflows for parametric modeling, drawings and engineering analysis. Requires FreeCAD and the relevant solver or platform dependencies. Generated geometry and simulation convergence are not engineering certification; the interface can execute Python and write files.