# Dare MCP Server [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/avi-aggarwal14/dare-mcp-server  
**GitHub Stars:** 1  
**npm Downloads (last month):** 38  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/avi-aggarwal14-dare-mcp-server

## Description
Generate video and images on Dare (Seedance 2.5 and 2.0, Veo 3.1, Kling 3.0 Pro, Hailuo 2.3, Nano Banana 2, GPT Image 2, Seedream 5 Pro) from Claude, billed to your own Dare credits, with local cost estimates and dry runs.

## Tools
Capabilities this server exposes over MCP:

- **dare_list_models** — Models with durations, qualities, aspect ratios, reference limits
- **dare_estimate_cost** — Local credit estimate before you commit
- **dare_get_credit_balance** — Current balance
- **dare_generate_video** — Text-to-video, Seedance 2.5 by default; optional reference media. `dry_run: true` previews the spec and price for free
- **dare_generate_image** — Stills via Nano Banana 2, GPT Image 2, Seedream 5 Pro. Supports `dry_run
- **dare_get_generation** — Poll status, get the output URL
- **dare_list_generations** — Paginated library listing
- **dare_cancel_generation** — Stop an in-flight job
- **dare_delete_generation** — Delete from the library (destructive)
- **dare_upload_media** — Upload a reference image/clip/audio, returns a `storage_key
- **dare_list_uploads** — Previously uploaded references
- **dare_delete_upload** — Delete an uploaded reference (destructive)
- **dare_list_projects** — Your Dare projects

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "dare-mcp-server": {
    "command": "npx",
    "args": ["-y","dare-mcp-server"],
    "env": {
      "DARE_CLIENT_TOKEN": "",
      "DARE_CONFIG_DIR": "",
      "DARE_MAX_CREDITS_PER_CALL": "",
      "DARE_UPLOAD_ROOTS": "",
      "DARE_ALLOW_URL_UPLOADS": "",
      "DARE_MAX_UPLOAD_BYTES": "",
      "DARE_SESSION_TOKEN": "",
      "DARE_SESSION_ID": ""
    }
  }
}
```

**Requires environment variables:** `DARE_CLIENT_TOKEN`, `DARE_CONFIG_DIR`, `DARE_MAX_CREDITS_PER_CALL`, `DARE_UPLOAD_ROOTS`, `DARE_ALLOW_URL_UPLOADS`, `DARE_MAX_UPLOAD_BYTES`, `DARE_SESSION_TOKEN`, `DARE_SESSION_ID` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What Dare MCP Server does

Dare MCP Server MCP server gives MCP clients access to Dare’s media-generation workflow. It can start video jobs using models such as Seedance 2.5, Seedance 2.0, Veo 3.1, Kling 3.0 Pro, and Hailuo 2.3. For still images, it supports Nano Banana 2, GPT Image 2, and Seedream 5 Pro.

The server also exposes account and library operations. An agent can inspect available models and their constraints, check the current Dare credit balance, estimate a request’s credit cost locally, retrieve completed outputs, list past generations, and inspect Dare projects. Video and image generation can run as a dry run first, returning the planned specification and price without spending credits.

## How it works

Authentication uses the long-lived `__client` cookie from a signed-in Dare browser session. The server exchanges that credential for a short-lived session token when making calls, following the same general sign-in flow as the Dare web application. Credentials can come from the setup-created configuration file or from the `DARE_CLIENT_TOKEN` environment variable.

Generation requests return jobs that can be checked with `dare_get_generation`. The result includes the output URL when processing finishes. Reference images, clips, or audio can be uploaded first; the upload tool returns a `storage_key` that can be supplied to a generation request.

## Setup and configuration

Dare MCP Server MCP server requires Node.js 20 or newer. Running `npx dare-mcp-server setup` starts an interactive wizard, validates the supplied Dare credential, displays the balance, stores configuration under `~/.dare-mcp/config.json`, and can add the server to Claude Code or the Claude desktop app. The package also provides `npx dare-mcp-server check` for credential and service checks.

The default process uses stdio. `npx dare-mcp-server http` starts the optional HTTP transport for self-hosting. A source checkout can be built with npm and launched by pointing an MCP client at its `dist/cli.js` entry point.

Useful controls include `DARE_MAX_CREDITS_PER_CALL` for a per-generation credit limit, `DARE_UPLOAD_ROOTS` for permitted local upload directories, `DARE_ALLOW_URL_UPLOADS` for public URL uploads, and `DARE_MAX_UPLOAD_BYTES` for upload size limits. `DARE_CONFIG_DIR` changes the configuration location, while `DARE_REQUEST_TIMEOUT_MS` controls request timeouts.

## Tools and capabilities

- `dare_list_models`: Lists models, durations, qualities, aspect ratios, and reference limits.
- `dare_estimate_cost`: Calculates a local credit estimate before submission.
- `dare_get_credit_balance`: Reads the current Dare balance.
- `dare_generate_video` and `dare_generate_image`: Submit video or still-image jobs, with dry-run support.
- `dare_get_generation`: Polls a job and returns its output URL.
- `dare_list_generations`: Lists library items with pagination.
- `dare_cancel_generation`: Stops an active generation.
- `dare_delete_generation`: Removes a generation from the library.
- `dare_upload_media`, `dare_list_uploads`, and `dare_delete_upload`: Manage reference media.
- `dare_list_projects`: Lists Dare projects.

These capabilities place the Dare MCP Server MCP server in the Multimedia Process category, particularly for agent-assisted image and video creation workflows.

## Limitations and notes

Dare does not publish a developer API for this workflow. The server communicates with a private endpoint used by the Dare web application, so Dare may change the contract without notice and use may fall outside Dare’s terms of service. The browser credential can access the Dare account and spend credits; it should be treated like a password.

Generation and deletion actions have different consequences. Dry runs do not spend credits, but submitted media jobs do. Deleting generations or uploaded references is destructive. Local file uploads are refused unless their paths are allowed through `DARE_UPLOAD_ROOTS`; public URL uploads can be disabled, and private or link-local addresses are refused.

_Full upstream README: https://allmcps.com/mcp/avi-aggarwal14-dare-mcp-server/readme_