@elvenvtt/mcp-account

MCP server exposing Elven creator account management to AI agents (Claude Code, Cursor, Claude Desktop, hosted Claude, and anything else that speaks the Model Context Protocol).

The agent-facing companion to @elvenvtt/cli. Same auth, same verbs, different surface: tools instead of commands.

Full docs: docs.elvenvtt.com/docs/agent-platform/mcp-setup

Two transports — pick the one that fits your agent host:

Transport 1: stdio (Claude Code / Cursor / Claude Desktop)

For local agents that can spawn subprocesses.

npm install -g @elvenvtt/mcp-account

Add to your agent's MCP config (e.g. .mcp.json in your project root, or Claude Desktop's claude_desktop_config.json):

{
  "mcpServers": {
    "elven-account": {
      "command": "elven-mcp-account"
    }
  }
}

Transport 2: HTTP (cloud agents, web Claude, hosted Cursor, CI)

For agents that can't spawn subprocesses. Streamable HTTP, JSON-RPC over POST. No install needed — point your MCP client at the hosted endpoint.

{
  "mcpServers": {
    "elven-account": {
      "type": "http",
      "url": "https://mcp.elvenvtt.com/mcp",
      "headers": { "Authorization": "Bearer el_pat_..." }
    }
  }
}

Every request is stateless and re-verifies the token — no sessions, no cookies. The HTTP transport intentionally omits publish.companion (it reads .elven.json from a local project dir, which doesn't exist over HTTP); use the stdio transport or elven publish from a terminal when you need that flow.

Auth (stdio transport)

You must have either:

• Logged in via the CLI first — npx @elvenvtt/cli login. The stdio server reads the cached session from your OS keyring (or ~/.elven/credentials on Unix).
• OR set ELVEN_TOKEN to a PAT (el_pat_…) — best for CI / headless / cloud-hosted agents:

{
  "mcpServers": {
    "elven-account": {
      "command": "elven-mcp-account",
      "env": { "ELVEN_TOKEN": "el_pat_..." }
    }
  }
}

Auth (HTTP transport)

Pass a PAT via the Authorization header per the config above. JWTs also work but PATs are recommended for non-interactive agents.

Get a token

• In the Elven Studio Developer page under the CLI Tokens tab, or
• From the CLI: elven tokens create --label "claude-code".

Tools

24 tools total. Descriptions in the actual tool registration include usage hints for the agent (when to prefer one tool over another, which ones are destructive).

Identity

| Tool | Purpose | |---|---| | whoami | Identify the authed account. Call FIRST to confirm identity. |

Listings

| Tool | Purpose | |---|---| | listing.list | Browse public listings. | | listing.mine | Your own listings including drafts. | | listing.show | Full detail for one listing id. | | listing.create | Create a new listing (raw — usually prefer publish.companion). | | listing.update | Patch listing fields. | | listing.delete | DESTRUCTIVE — permanent. |

Publish

| Tool | Purpose | |---|---| | publish.companion | Build + upload + sync — the "one command from edit to live" flow. | | publish.uploadFile | LOW-LEVEL single-file upload. |

Posts

| Tool | Purpose | |---|---| | post.new | Create a post (draft by default; publishing fires a follower fanout). | | post.list | List a user's posts (yours includes drafts). | | post.show | Full content of one post. | | post.update | Patch fields. | | post.delete | DESTRUCTIVE. |

Drops

| Tool | Purpose | |---|---| | drop.create | Create a time-windowed drop. | | drop.list | List a user's drops. | | drop.show | Drop + your claim/eligibility state. | | drop.update | Patch fields. | | drop.delete | DESTRUCTIVE. |

Followers

| Tool | Purpose | |---|---| | followers.count | Public follower count for any user. | | followers.following | Who the authed user follows. |

Analytics

| Tool | Purpose | |---|---| | analytics.overview | Followers + subscribers + drops + recent activity (one round-trip). | | analytics.sales | Sales totals, top listings, recent purchases. |

Tokens

| Tool | Purpose | |---|---| | tokens.list | List your PATs (no plaintext). Create/revoke are intentionally NOT exposed via MCP — those go through the CLI or Studio. |

Design

• Pure-function handlers (input, ctx) → output. Adding an HTTP transport later is wrapping these.
• Token priority: ELVEN_TOKEN env > CLI cached creds > clear error to stderr + exit 1.
• Same backend as the CLI — all calls go through @elvenvtt/api. If something works in elven, it works here.
• Destructive tools are flagged in their descriptions — agents are expected to confirm with the user before calling them.

Security

The Authorization header is the credential. Treat it like a password — ELVEN_TOKEN env var instead of inline in MCP config, gitignore your .mcp.json if you ever inline the literal value, audit + revoke unused tokens from Studio Developer → CLI Tokens. Full guidance at docs.elvenvtt.com/docs/agent-platform/security.

License

MIT.