FlowchartAI MCP

What FlowchartAI MCP MCP server does

FlowchartAI MCP MCP server exposes FlowchartAI diagram operations to AI assistants through the Model Context Protocol. It works with editable flowcharts rather than producing a static image only. An assistant can turn a written process description or a supplied image into a diagram, then request revisions using ordinary language.

The server also provides access to saved diagrams and their related records. Agents can list diagrams, read individual diagrams, inspect task history, and review version history. For downstream processing, diagrams can be exported as editable draw.io XML or JSON. Saved work can also be opened in the FlowchartAI web editor, where users can continue editing and render PNG or SVG output.

As a 📊 Data Visualization tool, it fits workflows that need a visual representation of business processes, software flows, project procedures, or other structured ideas.

How it works

Requests are sent from an MCP-compatible AI client to FlowchartAI. Generation and modification tasks use the account’s existing allowance or Credits. Per-task spending caps and account or API-key budgets can limit those operations, so an agent workflow should account for available capacity before repeatedly generating or revising diagrams.

The hosted service exposes a Streamable HTTP MCP endpoint. It supports OAuth with PKCE and scoped API keys for access control. The hosted version requires a FlowchartAI account. Once a diagram is saved, later requests can refer to it through the server’s diagram, task, and version-history capabilities.

Setup and configuration

For the hosted service, users need a FlowchartAI account and an MCP client that can connect to a Streamable HTTP endpoint. Authentication can use OAuth with PKCE or a scoped API key, according to the access method configured for the endpoint.

The repository is also self-hostable, but the supplied project documentation describes the web application setup rather than a separate MCP deployment procedure. A self-hosted installation requires Node.js 22.13 or newer, pnpm, and PostgreSQL. AI functionality requires an OpenRouter API key, while Google OAuth credentials are listed as required for authentication. Cloudflare R2 or AWS S3-compatible storage is optional for file uploads.

The documented self-hosted configuration includes DATABASE_URL, BETTER_AUTH_SECRET, GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, and OPENROUTER_API_KEY. Optional storage and email variables are available for those application features.

Tools and capabilities

FlowchartAI MCP MCP server supports these operations:

• Generate a flowchart from natural-language text.
• Generate a flowchart from an image.
• Revise an existing diagram with natural-language instructions.
• List saved diagrams.
• Read saved diagram contents.
• Inspect task history and version history.
• Export editable draw.io XML or JSON.
• Open saved diagrams in the web editor for continued editing and PNG/SVG rendering.

Limitations and notes

AI generation and editing consume the account allowance or Credits; they are not described as unlimited on the hosted service. Spending controls can apply at the task, account, or API-key level. Manual editing, saving, and downloading are described as free for the hosted web application, but that does not remove the documented limits on AI work.

The public repository is MIT licensed, while use of the hosted service requires an account. The repository also describes Excalidraw-based editing and additional application export formats, but the MCP-specific export formats identified here are draw.io XML and JSON.