FluxDesk
← All integrations

Integrations · AI agent access

MCP (Model Context Protocol)

Let an AI agent read your inbox, mark items read, submit daily updates, query tasks / feedback / projects, and act on requests — all over a single Bearer-authed JSON-RPC endpoint.

Connection

Transport
Streamable HTTP (POST /mcp)
URL
https://my.fluxdesk.net/mcp
Auth
Authorization: Bearer sk-flux-…
Issuance
Self-serve at /me/agent (any authenticated user) or admin issues via access_credentials with credentialType='ai_gateway_token'.

List available tools

Round-trip a standard JSON-RPC tools/list to discover the live tool set (38+ as of v0.19.x). New tools land here without you bumping any client lib.

shell
curl -X POST https://my.fluxdesk.net/mcp \
  -H "Authorization: Bearer sk-flux-EXAMPLE…" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Claude Code config

Drop into ~/.claude/config.json (or per-workspace). Restart Claude Code; tools appear under the fluxdesk namespace.

jsonc
{
  "mcpServers": {
    "fluxdesk": {
      "type": "http",
      "url": "https://my.fluxdesk.net/mcp",
      "headers": {
        "Authorization": "Bearer sk-flux-EXAMPLE…"
      }
    }
  }
}

Tool families (live count via tools/list)

  • Inbox

    list_inbox, get_notification, mark_*_read.

  • Daily check-in

    submit_daily_update + recall.

  • Tasks

    Create / dispatch / status / blocker.

  • Feedback

    Comment, edit, mark-resolved.

  • Projects

    List, read, dashboard summary.

  • Inbound delivery

    Resolve / supplement / lifecycle.

  • Outbound

    Manage your own webhooks.

  • Team-support

    Approval requests / catalog (admin).

Full guide