# Flowra [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/flowradev/skills  
**GitHub Stars:** 1  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/flowra-2

## Description
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows

## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: high). Install path detected from listing signals. Add as a URL/SSE server in your client:

```json
"mcpServers": {
  "flowra": {
    "url": "https://mcp.flowra.dev/mcp"
  }
}
```

## Documentation

## What Flowra MCP server does

Flowra MCP server exposes hosted tools for working with connected applications and custom tools through the Model Context Protocol. The documented app coverage includes Gmail, Slack, GitHub, Notion, Telegram, WhatsApp, Linear, and more than 1,000 apps. It can support hosted AI agents and approval-gated workflows, including scheduled runs.

The repository also includes agent skills and plugins for Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI, and VS Code. Those files help route requests to the appropriate Flowra calls and provide setup and status commands. The MCP endpoint itself is hosted at `https://mcp.flowra.dev/mcp`.

## How it works

A client connects to the hosted endpoint and selects a Flowra project. OAuth is used for normal account authentication, so an API key is not needed for the standard setup. After connecting an account, an agent can discover available capabilities, inspect schemas, and execute operations across the configured services.

Flowra MCP server separates read operations from actions that change data or send messages. Reads can run without confirmation, while sends, posts, creates, updates, deletes, and payments require an explicit approval. Scheduled workflows must retain an approval step through a `human_wait` node or an agent `humanApproval` setting.

Each tool call has a ledger entry containing its inputs, outputs, and credits. Inbound email, chat, and webhook content is treated as untrusted data rather than instructions.

## Setup and configuration

Install the repository skill with:

```bash
npx skills add flowradev/skills --skill flowra
```

Then register `https://mcp.flowra.dev/mcp` in the MCP client. Cursor can connect it from its MCP settings; Claude Code can add it over HTTP; VS Code can add the URL to `.vscode/mcp.json`; and Gemini CLI supports the repository extension from release `v0.2.0`.

For CI or clients without OAuth, a project API key and username can be supplied as local configuration headers. The README specifically warns not to commit those values to the repository. No environment variables are documented as required.

## Tools and capabilities

The hosted MCP registration exposes six tools:

- `discover` for finding available tools
- `schemas` for inspecting tool definitions
- `connect` for connecting an account
- `multi-execute` for executing multiple operations
- `workbench` for Flowra workbench operations
- `bash` for bash-related execution

The repository also provides `/flowra-setup` and `/flowra-status` commands, onboarding instructions, safety rules, and behavioral evaluations. Custom REST endpoints can be described to Flowra so it can build a toolkit for an app that is not already listed.

## Limitations and notes

A Flowra project is required, and the documented use cases assume a hosted Flowra setup. The README recommends skipping it when the task is already being managed as an n8n, Zapier, or Make graph. OAuth is the default path; API-key configuration is intended for CI or clients that cannot use OAuth. Approval gates remain relevant for durable automations and scheduled actions.

_Full upstream README: https://allmcps.com/mcp/flowra-2/readme_