FlowDot is an AI workflow automation platform that enables users to build, share, and run AI workflows using a visual drag-and-drop interface. It supports multiple AI providers including ChatGPT, Claude, and Gemini, and offers features like MCP integration, built-in knowledge base with RAG, and energy usage transparency.

What is the difference between a workflow and a recipe?

A workflow is a visual node graph that runs on the FlowDot server and is designed for quick automation. A recipe is an agentic orchestration program that runs on the local FlowDot CLI for longer tasks with loops, branches, and human approval gates.

Can I use FlowDot with Claude Desktop or Cursor?

Yes. FlowDot provides an MCP server that gives AI clients platform access across workflows, apps, custom nodes, recipes, knowledge base, toolkits, and execution controls.

Do I have to use FlowDot credits?

No. You can route models through FlowDot credits or connect your own provider keys using BYOK, depending on how you want to manage access and billing.

Visual workflows. Local agents. MCP-native.

Build once. Run across the whole AI stack.

FlowDot is the AI automation platform for teams that need more than a drag-and-drop workflow tool. Design server-side workflows, run long local CLI recipes, connect Claude and Cursor through MCP, ship React apps, and route models through FlowDot credits or your own keys.

Server workflows Fast visual automations, custom nodes, scheduled runs, and sharable execution results.

Local recipes Long-running agentic tasks with loops, branches, approvals, and real code actions.

MCP access Expose the platform to Claude Desktop, Cursor, Claude Code, and other MCP clients.

Apps and toolkits Turn workflows into products with React frontends, reusable tools, and credential-aware integrations.

Create Your Workspace See the Docs Explore MCP

80+ MCP tools Free tier with 5 executions/day FlowDot credits or BYOK Knowledge base + RAG Apps, workflows, recipes, toolkits

Search public workflows, apps, docs, and tutorials.

Workflows Recipes Apps API Reference Compare FlowDot

The FlowDot ecosystem

One platform, multiple runtimes

Web Editors

Workflows, apps, recipes, toolkits, and custom nodes are designed in-browser.

Server Workflow Runtime

Quick automations execute on the platform with scheduling, sharing, and team access.

AI Clients MCP Server

Claude Desktop and Cursor can reach the platform through MCP.

Local CLI Recipes

Long-running agent loops execute locally with approvals and file or shell actions.

Core FlowDot Hub

Laravel-wrapped platform coordinating the React client, Node execution layer, auth, and routing.

Product React Apps

Ship frontends that invoke linked workflows without leaving the platform.

Knowledge RAG + Teams

Upload docs, query knowledge, and share resources across team workspaces.

Integrations Toolkits

Package HTTP tools, workflow tools, and credentials into reusable agent capabilities.

Models Credits or BYOK

Use FlowDot routing or direct provider keys without locking your stack to one vendor.

Server Visual workflows

Node-based graphs for AI, HTTP, custom logic, scheduling, and sharable outputs.

Local CLI recipes

Agentic orchestration that can run for minutes or hours with human gates.

MCP 80+ tools

Expose workflows, apps, nodes, recipes, knowledge, and executions to AI clients.

Apps React frontends

Convert internal automations into user-facing products powered by linked workflows.

Models Credits or BYOK

Keep routing flexible across providers, tiers, and user-owned API keys.

Platform Surfaces

FlowDot is not one feature. It is a full operating layer for AI work.

Most products stop at a workflow canvas. FlowDot extends beyond that into local agent execution, MCP exposure, app shipping, reusable toolkits, and model routing that works for both centralized teams and BYOK users.

Workflows on the server

Build fast visual automations that execute on the platform and can be triggered from the web UI, API, CLI, or MCP.

Visual graph editor: Connect AI, HTTP, input, output, and custom node logic without writing application code.
Execution anywhere: Run workflows from browsers, public pages, APIs, scheduled tasks, or AI clients.
Reusable artifacts: Share workflows publicly, track executions, and create presets or community inputs.

Recipes on the local CLI

Recipes handle the long-running agentic work that should happen close to the codebase, terminal, and human operator.

Long-form runtime: Plan, branch, loop, parallelize, and pause for approval when the task needs real supervision.
Real local actions: Search code, read files, edit content, and execute shell commands where the work actually lives.
Designed on web, cooked locally: Define recipes in the web editor and execute them through the FlowDot CLI.

Apps, toolkits, and custom nodes

Turn automations into products and reusable capabilities instead of leaving them trapped inside one-off flows.

React apps: Build frontends that invoke linked workflows and expose polished interfaces for internal or external users.
Agent toolkits: Bundle HTTP tools, workflow tools, and credential mappings into installable integrations.
Custom nodes: Extend the workflow runtime with specialized logic and optional LLM-backed behavior.

Knowledge, access, and model routing

Bring documents, provider access, teams, and outside AI clients into the same operating model.

Knowledge base: Upload documents, query them with RAG, and organize access across personal or team scopes.
MCP access: Let Claude Desktop, Cursor, and other compatible clients operate on the platform directly.
Flexible model access: Use FlowDot credits, BYOK, tiered preferences, and transparency features in one place.

Execution Model

The key distinction: workflows and recipes solve different classes of work.

That split is part of the platform's strength. Quick automations stay on the server. Longer agentic tasks stay local where they can touch code, files, shells, and approval gates safely.

Workflow

Fast, visual, server-side automation

Use workflows when you want a durable automation graph that can be shared, scheduled, embedded, or executed on demand.

Runs on: FlowDot server
Best for: AI pipelines, APIs, forms, shared business processes, public tools, reusable automations
Invocation: Web UI, API, CLI, MCP, schedules
Typical duration: Seconds to minutes

Recipe

Long-running, local, agentic orchestration

Use recipes when the job behaves more like a working session than a single automation run.

Runs on: Local FlowDot CLI
Best for: Code review, repo-wide analysis, content production, research loops, multi-step agent tasks
Invocation: CLI runtime
Typical duration: Minutes to hours with approvals and branches

Model Access

Use the platform's model routing when it helps. Bring your own keys when that fits better.

FlowDot is designed to avoid forcing a single provider strategy. Teams can standardize through FlowDot credits, while advanced users can still plug in direct provider access and preferred tiers.

FlowDot Credits Path

Centralized routing for speed and consistency

Choose FlowDot when you want a unified access layer across providers, model tiers, and internal routing.

FlowDot -> Routing Provider -> Source Provider -> Model

Quick-select tiers: Simple, capable, and complex model preferences can stay consistent across the platform.
Aggregation layer: Route across providers without making end users manage every upstream detail.
Visibility: Pair usage with transparency features, including environmental impact reporting.

BYOK Path

Direct provider control when you need it

Use your own provider keys if billing, security, or model access needs to stay in your hands.

Your API Key -> Direct Provider API -> Selected Model

No lock-in: Switch between FlowDot-managed access and direct provider access by use case.
Provider flexibility: Connect preferred LLM or audio providers through the same platform surface.
User-level control: Configure model access in one place instead of hardcoding provider choices into each project.

Start with a workflow. Expand into a platform.

Most teams enter through one automation. The upside comes when that workflow becomes a shared runtime, a local recipe, an MCP tool surface, or a customer-facing app. FlowDot is built to let that expansion happen without re-platforming.

Start Free Why FlowDot Compare Platforms

FAQ

Questions teams ask before they commit

These are the questions that usually matter when deciding whether FlowDot is just another automation tool or an actual AI operating layer.

What is FlowDot in one sentence?

FlowDot is an AI automation platform that combines server-side workflows, local CLI recipes, MCP-connected AI clients, React apps, toolkits, custom nodes, and knowledge base retrieval in one system.

Why not just use a workflow builder?

Because many real tasks are not only workflows. Some need long-running local agent loops, some need MCP access for Claude or Cursor, some need reusable toolkits, and some need productized frontends. FlowDot covers that entire span instead of stopping at the graph editor.

Can non-developers still use it?

Yes. The platform starts with visual editors and public pages. Developers can go deeper with custom nodes, recipes, MCP, and app code, but the entry point does not require writing application logic.

Can I start free and keep provider flexibility?

Yes. FlowDot includes a free tier with 5 workflow executions per day, and the platform supports both FlowDot-managed access and BYOK model configuration depending on how you want to operate.