VORDIXRequest a demo
Back to home

Self-host your MCP gateway on your infrastructure.

Deploy Vordix on infrastructure you control. Connect MCP and REST clients to one gateway, keep governance records in your deployment and choose which upstream tools and AI providers receive data.

Plan your self-hosted MCP gateway deployment

For a pilot, the documented starting point is a Linux host with Docker and Docker Compose v2. Production planning also needs DNS, TLS, persistent storage and a tested backup and restore process.

  1. Prepare the environment

    Choose a host under your control and a DNS name for the portal. Configure TLS for production and outbound connectivity to the upstream tools you intend to connect.

  2. Configure and start Vordix

    Use the supplied deployment configuration to set up the application, database, Redis and integration credentials. Keep secrets out of source control and use persistent storage for stateful services.

  3. Start with one integration

    Connect a tool such as Jira, restrict the operations and resources, issue a caller key and configure the MCP client. Test denied requests and response policies before onboarding more teams.

  4. Make the data boundary explicit

    Self-hosting controls where the gateway runs. Requests still reach configured upstream services, and returned content may reach your selected AI provider. Review both alongside retention and backup requirements.

Why independence matters

Governance you don’t control isn’t really governance, it’s just another dependency.

  • 01

    SaaS governance tools see your data

    A cloud governance layer means your toolchain traffic and PII pass through a third party’s network, a non-starter under EU and DACH residency rules.

  • 02

    AI-vendor lock-in

    Build your controls around one provider’s connector and audit tooling, and you re-plumb everything the day you switch models.

  • 03

    A black-box runtime breaks accountability

    Containerised “agent-in-a-box” runtimes reintroduce exactly the opacity governance is meant to remove.

  • 04

    Generic connectors over-trust the tool

    A thin pass-through that forwards raw SQL or raw queries can neither contain injection nor scope access.

Works with any AI tool, without lock-in, on your infrastructure.

An external gateway in front of all your tools is what auditors trust, and it keeps you free to switch AI providers at any time.

Works with any AI tool

Connect Claude, Cursor, VS Code, an internal chat tool, or anything else that speaks MCP or REST.

No vendor lock-in

Switch AI providers without re-plumbing your governance, because the gateway stays the same.

Self-hosted on your infrastructure

Runs in your own network via Docker Compose, and Vordix as a company is never in the request path, so your data never leaves.

Works with any MCP or REST client
  • Claude
  • Claude Code
  • Cursor
  • VS Code
  • Internal chat tools

How Vordix solves it

  • 01

    Self-hosted on your infrastructure

    Runs entirely in your own network via Docker Compose, one deployment per organisation, not a multi-tenant SaaS.

  • 02

    No third-party in the data path

    There is no Vordix-operated cloud and no AI-vendor hop; your credentials, traffic and audit logs never leave your network, and PII detection runs offline.

  • 03

    One gateway for any AI client

    Point Claude, Claude Code, Cursor, VS Code or an internal chat tool at the same governed endpoint, and switch providers without re-plumbing your governance.

  • 04

    Works over MCP or plain REST

    Standard Authorization: Bearer auth, an OpenAPI schema for REST clients and an MCP manifest for everything else.

  • 05

    Every integration is a governance layer

    Native adapters, not pass-throughs. Databricks accepts no raw SQL, for instance, so injected values are inert against the real database.

  • 06

    You own the keys and the backups

    Credentials are encrypted with AES-256-GCM, the encryption keys live only in your .env, and backups are encrypted nightly.

  • 07

    Route the AI provider too In progress

    Send the model call itself through Vordix, with OpenAI and Anthropic adapters behind an OpenAI-compatible endpoint, so you can swap providers freely.

  • 08

    Governed agentic workflows Planned

    A deterministic workflow builder with agent nodes, scheduling and cost tracking, where the graph itself is the governance surface.

See it on your own stack.

A short walkthrough on your tools, your rules, your audit log. Nothing leaves your network.

Request a demo