DeployStack - Open Source MCP Hosting Platform

<p align="center">
<a href="https://deploystack.io">

</a>
</p>

<p align="center">
<a href="https://opensource.org/licenses/AGPL-3.0"></a>
<a href="https://github.com/deploystackio/deploystack/stargazers"></a>
</p>

<p align="center">
<a href="https://cloud.deploystack.io">🚀 Try it out</a> ·&nbsp;
<a href="https://deploystack.io">🌐 Website</a> ·&nbsp;
<a href="https://docs.deploystack.io/">📚 Documentation</a> ·&nbsp;
<a href="https://github.com/deploystackio/deploystack?tab=readme-ov-file#roadmap"> Roadmap </a> ·&nbsp;
<a href="https://discord.gg/42Ce3S7b3b"> Discord</a>
</p>

---

Deploy MCP servers from GitHub to HTTP endpoints in 30 seconds. Works with n8n, Dify, Voiceflow, Langflow, Claude Code, Cursor, and any MCP client.

The Problem

Most MCP servers are stdio-only. But workflow automation platforms need HTTP endpoints.

| Platform | What They Need | The Problem |
| ------------- | ----------------------------- | --------------------------- |
| n8n | HTTP URLs for MCP Client node | stdio doesn't work |
| Dify | HTTP/SSE endpoints (v1.6+) | No stdio support |
| Voiceflow | Remote MCP endpoints | "Local MCP NOT supported" |
| Langflow | HTTP/SSE for production | Local processes won't scale |

Current workarounds are painful:

- mcp-remote requires local Node.js setup
- Docker containers need DevOps expertise
- DIY hosting (Fly.io, Railway) takes hours

And for teams, the problems multiply:

- Credentials scattered across Slack DMs and .env files
- No visibility into who uses which tools
- Onboarding new developers takes hours
- When someone leaves, good luck rotating all those API keys

Who DeployStack Is For

For AI Automation Builders

- n8n, Dify, Voiceflow, Langflow users: Get HTTP endpoints for MCP servers - no more stdio limitations
- No DevOps required: Skip the Docker containers and mcp-remote setup
- Deploy your custom MCP servers: Point at a GitHub repo, get a URL in 30 seconds
- Focus on workflows: Build automations, not infrastructure

For Automation Agencies

- Client-ready MCP hosting: Deploy MCP servers for client projects without managing infrastructure
- White-label ready: Team isolation keeps client data separate
- Scale without ops burden: Add more MCP servers without adding DevOps headcount
- Credential management: Secure vault for all your clients' API keys

For Platform / DevOps Teams

- End credential chaos: Move API keys from Slack DMs and .env files into an encrypted vault with automatic injection
- Control MCP access: Role-based permissions determine who can use which MCP servers and tools
- Prevent data exposure: Team isolation and audit logging show exactly what tools accessed what data
- Centralize configuration: One source of truth instead of N developers × M configurations

For Developers

- Zero setup: Add one URL to your MCP client, get access to all team MCP tools instantly
- No local installs: No npm install, no process management, no port conflicts
- Just works: Same configuration everywhere - no more "works on my machine" debugging
- Focus on code: Security and credentials are handled for you

For Engineering Managers

- 1-click team onboarding: New developer? They're productive with MCP tools in minutes, not hours
- Reduce AI costs: 98% token reduction means lower API bills and better LLM performance
- Full visibility: See which tools your team uses, how often, and what they access
- Offboarding safety: When someone leaves, revoke access instantly - no credential rotation nightmare

The Solution

DeployStack hosts MCP servers as HTTP endpoints. Two ways to use it:

Deploy from GitHub

Point DeployStack at any GitHub repo with an MCP server:

1. Connect your GitHub account
2. Select a repository
3. Get an HTTP endpoint that works immediately

Auto-deploys when you push. Like Vercel, but for MCP servers.

Use the Curated Catalog

Don't have your own MCP server? Browse our catalog of popular servers:

- One-click installation
- Pre-configured and tested
- Ready to use in seconds

Key Features

Deploy from GitHub

GitHub Repo → DeployStack → HTTP Endpoint → Your MCP Client

- Any stdio MCP server becomes an HTTP endpoint
- Auto-deploy on git push
- Branch deployments for testing
- Zero DevOps required

98% Token Reduction

Traditional MCP setup loads all tools into context:

- 10 servers × 15 tools = 150 tools = 75,000 tokens consumed

DeployStack's hierarchical router exposes only 2 meta-tools:

- discover_mcp_tools - find relevant tools
- execute_mcp_tool - run the tool you need
- Result: 1,372 tokens instead of 75,000

Scale to 100+ MCP servers without degrading LLM performance.

Credential Vault

Store API keys and secrets securely:

- Encrypted at rest
- Automatic injection at runtime
- No credentials in config files
- No secrets in Slack DMs
- Rotate once, update everywhere

Team Management & RBAC

Control access across your team:

- Invite team members with specific roles
- Define who can use which MCP servers
- Manage permissions from one dashboard
- Instant onboarding for new developers

Audit Logging

Full visibility into MCP usage:

- Track every tool interaction
- See who accessed what and when
- Export logs for compliance
- Debug issues with complete history

Works Everywhere

Compatible with any MCP client that supports HTTP/SSE:

| Workflow Platforms | AI Coding Tools | Other Clients |
| ------------------ | --------------- | ------------------- |
| n8n | Claude Code | Any HTTP MCP client |
| Dify | Cursor | Custom integrations |
| Voiceflow | VS Code | |
| Langflow | Windsurf | |

Architecture

┌─────────────┐     ┌─────────────────┐     ┌─────────────┐
│   GitHub    │────▶│   DeployStack   │────▶│  HTTP URL   │
│    Repo     │     │    Satellite    │     │   Endpoint  │
└─────────────┘     └─────────────────┘     └─────────────┘
                            │
                    ┌───────┴───────┐
                    │               │
              ┌─────▼───-──┐   ┌───-──▼──-───┐
              │ Credential │   │ Hierarchical│
              │   Vault    │   │   Router    │
              └──────────-─┘   └─--──────────┘

Two deployment options:

| Global Satellites | Team Satellites |
| -------------------------- | ----------------------------- |
| Managed by DeployStack | Deploy on your infrastructure |
| Zero setup | Full data control |
| Free tier available | Enterprise security |
| satellite.deploystack.io | satellite.yourcompany.com |

Quick Start

1. Sign Up

Create a free account at cloud.deploystack.io

2. Add MCP Servers

Option A: Deploy from GitHub

- Connect your GitHub account
- Select a repository with an MCP server
- Click Deploy

Option B: Use Catalog

- Browse the MCP server catalog
- Click Install on any server
- Configure credentials if needed

3. Get Your Endpoint

Copy the HTTP endpoint URL from your dashboard.

4. Configure Your MCP Client

Add to your MCP client configuration:

{
  "mcpServers": {
    "deploystack": {
      "type": "http",
      "url": "https://satellite.deploystack.io/mcp"
    }
  }
}

That's it. All your MCP servers are now accessible via one URL.

Self-Hosting

DeployStack is open source (AGPL-3.0). Run it on your own infrastructure:

git clone https://github.com/deploystackio/deploystack
cd deploystack
docker-compose up

See Self-Hosting Documentation for detailed instructions.

Roadmap

Phase 1: Foundation (Completed)

- [x] Deployed cloud.deploystack.io hosted version with backend and frontend
- [x] Implemented secure user and team management system with roles and permissions
- [x] Integrated OAuth for secure logins (GitHub)
- [x] Created initial MCP Server Catalog for tool discovery
- [x] Established documentation and self-hosted Docker support
- [x] PostgreSQL-only database architecture for enterprise scalability

Phase 2: Enterprise Governance (Completed)

- [x] Auto-install MCP servers for new users with admin-controlled defaults
- [x] Featured MCP servers filtering for improved tool discovery
- [x] Global Event Bus System with plugin integration
- [x] Three-Tier MCP Configuration Architecture (template, team, user levels)
- [x] Multi-User Configuration Management with personalized device-specific configs
- [x] Advanced MCP Argument & Environment Variable Handling with schema validation
- [x] HTTP Headers & URL Query Parameters Support for HTTP/SSE transport
- [x] Selective Secret Masking in Satellite Logs

Phase 3: Satellite Infrastructure (Completed)

- [x] Global Satellite Infrastructure - managed MCP servers via HTTPS
- [x] Zero-Installation Experience - just add URL to your MCP client
- [x] OAuth Authentication with token-based auth
- [x] Satellite Pairing Security with JWT-based token registration
- [x] Public Launch - production satellite for community use
- [x] Resource Management - process isolation and limits
- [x] GitHub README and Stars Integration for MCP catalog
- [x] Background Job Queue System with PostgreSQL-based worker infrastructure
- [x] Cron Job Scheduling System with node-cron integration
- [x] MCP Registry Integration - official MCP Registry support

Phase 4: Context Window Optimization (Completed)

- [x] Hierarchical Router Implementation - 2-tool pattern achieving 98% token reduction
- [x] Semantic Tool Search with Fuse.js fuzzy matching (2-5ms performance)
- [x] Tool Discovery API with paths, descriptions, transport types, relevance scores
- [x] Smart Tool Execution with automatic routing to stdio or HTTP/SSE backends
- [x] MCP Tool Metadata Collection with token counting and event-driven sync
- [x] Tool Display & Analytics API for team-aware tool management
- [x] Token Analytics Dashboard with visual comparisons and per-installation breakdown
- [x] OAuth 2.1 Authentication for MCP Servers (Box, Linear, GitHub Copilot, etc.)

Phase 5: Advanced Governance (Completed)

- [x] Analytics dashboards for tool usage and performance
- [x] Per-tool access controls - team admins can disable individual tools
- [x] Improved searchable MCP Server Catalog with Featured page and categories
- [x] AI Instructions - ready-to-copy instruction files for AI coding assistants

Phase 6: Advanced Architecture (Completed)

- [x] Multi-Transport Support - SSE, Streamable HTTP, Direct HTTP protocols
- [x] Real-Time Command Orchestration with instant status feedback
- [x] Satellite Job System with JobManager and extensible architecture
- [x] Satellite Backend Events System with batch processing and JSON schema validation
- [x] MCP Client Activity Tracking across all components
- [x] Time-Series Metrics System with 15-minute bucket aggregation
- [x] Multi-region satellite deployment

Phase 7: GitHub Deployment (Completed)

- [x] Deploy MCP servers from any GitHub repository
- [x] Auto-deploy on git push via webhooks
- [x] Branch deployments for testing
- [x] Team Satellites - customer-deployed satellites for enterprise

Phase 8: In Progress

- [ ] Advanced Audit UI - detailed logging dashboard
- [ ] Python MCP Server Support - deploy Python-based MCP servers

Phase 9: Planned

- [ ] SSO Integration (SAML, OIDC)
- [ ] SOC 2 Type II certification
- [ ] AI-powered tool recommendations based on conversation context
- [ ] Per-team token budgets and cost attribution

Contributing

We welcome contributions. Here's how:

1. Fork the repository
2. Create a feature branch: git checkout -b feature/your-idea
3. Follow TypeScript and ESLint conventions
4. Add tests for new functionality
5. Submit a pull request

Areas where we need help:

- MCP server integrations
- Documentation improvements
- Satellite infrastructure
- Enterprise features

Links

- Platform: cloud.deploystack.io
- Documentation: docs.deploystack.io
- GitHub: github.com/deploystackio/deploystack
- Discord: Join the community

License

AGPL-3.0 - you can self-host, modify, and use DeployStack freely. If you offer it as a service, you must open-source your modifications.