mcpField
D
daemoon

daemoon

Remote
daemoon.devCloud & Infra0Updated 18w ago

About daemoon

daemoon is an MCP gateway for AI coding agents that centralizes access to connected development infrastructure providers. Users connect providers once and give agents a single token instead of storing provider secrets in each project.

Key features

  1. 01Unified tools for connected providers
  2. 02Browser-based provider connection
  3. 03Single MCP token for agents
  4. 04Encrypted provider-token vault
  5. 05Per-provider disconnect and token revocation
  6. 06Open-source self-hosting

Use cases

  1. 01Deploying Vercel projects from a coding agent
  2. 02Managing GitHub repositories and branches
  3. 03Running Supabase SQL through an agent
  4. 04Accessing several infrastructure providers with one token

MCP support

daemoon exposes provider operations as tools through an MCP server at `/api/mcp`, using Streamable HTTP and OAuth 2.0. MCP clients authenticate with a bearer token or supported OAuth flow.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://daemoon.dev/api/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://daemoon.dev/api/mcpAuth required

Tools

Details

From the project README

Daemoon

One login. Every infrastructure. Built for AI coding agents.

daemoon.dev · MCP server · Vercel · GitHub · Cloudflare · Supabase · GCP


Daemoon is the simplest way to give your AI coding agent (Claude Code, Cursor, etc.) access to all your dev infrastructure. Connect each provider once in your browser, then any agent gets unified, secure tool access via a single MCP token.

Why

You're prototyping with Claude Code. The agent wants to deploy a Vercel project, create a Supabase table, push a GitHub branch. Today you paste five different tokens into five different config files — and every new agent / repo / teammate starts over.

Daemoon ends that. One sign-in connects everything. One token wires the agent.

[You]                                    [Daemoon]                   [Providers]
  │                                          │                            │
  ├── Sign in once (Google) ─────────────────▶                            │
  ├── Connect Vercel · GitHub · CF · Supabase ▶─── OAuth · PAT ───────────▶
  ├── Generate MCP token ────────────────────▶                            │
  │                                          │                            │
  ▼                                          ▼                            ▼
 Claude Code / Cursor / any MCP client ⇄ /api/mcp ⇄ vault ⇄ live provider APIs

Install

Generate your token once at daemoon.dev/dashboard, then point your agent at https://daemoon.dev/api/mcp with header Authorization: Bearer dmn_....

Claude Code

Paste this in the Claude Code chat — Claude runs the install:

Install the Daemoon MCP server for me by running:
claude mcp add --transport http daemoon https://daemoon.dev/api/mcp --header "Authorization: Bearer dmn_..."
Cursor

Settings → MCP → Add server → HTTP:

  • URL: https://daemoon.dev/api/mcp
  • Headers: Authorization: Bearer dmn_...
Cline / Continue / Zed / any MCP client

Add an HTTP MCP server pointing at https://daemoon.dev/api/mcp with the bearer token header. Supports MCP Streamable HTTP (protocol 2025-06-18) and OAuth 2.0 (DCR + PKCE) for clients that prefer it.

Machine-readable
  • https://daemoon.dev/.well-known/mcp.json — server card
  • https://daemoon.dev/llms.txt — agent-readable summary
  • https://daemoon.dev/.well-known/oauth-protected-resource — RFC 9728
  • https://daemoon.dev/.well-known/oauth-authorization-server — RFC 8414

Connected providers

ProviderAuthSample tools
VercelOAuth (1-click)vercel.list_projects, vercel.create_deployment
GitHubOAuth (1-click)github.list_repos, github.create_repo
SupabaseOAuth + PATsupabase.list_projects, supabase.run_sql
CloudflarePATcloudflare.list_zones, cloudflare.create_dns_record
Google CloudService Account JSONgcp.list_projects, gcp.list_services
StripePAT (Secret key)stripe.list_products, stripe.list_customers, stripe.create_payment_link
ResendPATresend.send_email, resend.list_domains
OpenAIPATopenai.list_models, openai.chat_completion
AnthropicPATanthropic.create_message
npmPAT (Automation / Granular)npm.whoami, npm.list_packages, npm.view

More providers (Linear · Sentry · PostHog · Neon · Upstash · …) on the roadmap. PRs welcome.

How it's secure

  • Envelope encryption — every provider token is wrapped with a per-token DEK, then by a master key. AES-256-GCM end to end.
  • No raw secrets in your repo, no env vars to leak. Tokens live in the Daemoon vault; the agent never sees them, only the responses.
  • Revoke any token in one click. Per-provider disconnect, per-agent PAT revoke.

Self-host

Daemoon is open-source. Clone, set env vars, deploy to your own Vercel:

git clone https://github.com/daemoon-dev/daemoon
cd daemoon
cp .env.local.example .env.local   # add your Supabase + provider OAuth creds
npm install
npm run dev

Schema in sql/0001_init.sql.

Architecture

  • Next.js 15 App Router on Vercel (icn1 region).
  • Supabase for auth (Google + magic link) + Postgres vault.
  • MCP HTTP transport at /api/mcp — JSON-RPC tools/list + tools/call.
  • Connector interface in lib/connectors/types.ts: every provider implements OAuth / PAT validation + a list of tool defs. Adding a provider is one file.

Roadmap

  • More connectors: Stripe, Resend, Linear, Notion, Plausible
  • 1-click GitHub OAuth verified app (currently in Google Cloud "Testing" mode)
  • Service Account JSON → OAuth for GCP (when verified)
  • npm shim daemoon-mcp for stdio MCP clients

License

MIT. See LICENSE.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.