mcp.run
What is mcp.run?
Mcp.run is an enterprise AI connectivity platform for IT teams that runs a standards-compliant MCP gateway and management layer between apps and AI tools. It combines The Whole Org Chart, Audit & Control, and Advanced Security to govern access across internal systems. The platform is self-hosted and cloud-ready, integrates with OIDC-compatible identity providers, and includes Continuous Updates, Audit Usage, and 1-click Deactivation.
Last verifiedHow we evaluate
At a glance
- Mcp.run is best for enterprise IT teams who need governed AI access across many internal apps.
mcp.run now redirects to a different product
mcp.run was a hosted registry of "servlets" (portable, sandboxed WebAssembly tools) that AI agents could call over MCP, built by Dylibso — the company behind the Chicory WebAssembly runtime (1,135 GitHub stars) and the Extism plugin framework. The Internet Archive shows the real mcp.run web app still running on 7 September 2025 (https://web.archive.org/web/20250907083105/https://www.mcp.run/). By 1 February 2026 the same URL returns an HTTP 301 to turbomcp.ai, and it still does as of the most recent capture on 13 September 2026 (https://web.archive.org/web/20260913181258/https://mcp.run/). We confirmed the redirect is live today: fetching https://mcp.run/ now serves Turbo MCP's homepage directly. Turbo MCP is a different product aimed at a different buyer — a self-hosted MCP gateway for enterprise IT, not a public tool registry.
What Turbo MCP is, per the vendor
Turbo MCP describes itself as a "standards-compliant, self-hosted MCP gateway and management platform" for organizations, not individual developers (https://turbomcp.ai/). Per its features page, it centralizes an org's MCP servers behind one URL with automatic OAuth/Dynamic Client Registration, gives admins an approval workflow for which servers teams can use (RBAC-gated registration and testing before approval), and adds an audit log with a "kill-switch" to cut an agent's access to data without disrupting the underlying human account (https://turbomcp.ai/features). It ships as a single-container deployment meant to run on Kubernetes, a PaaS, or VMs behind the customer's own firewall — the pitch is that agent traffic never leaves the buyer's infrastructure.
No public pricing, no self-serve signup
Every primary call to action on turbomcp.ai — "Run Turbo MCP," "Book a Demo" — routes to a contact form or demo-booking page; there is no plan list, no per-seat or per-server number anywhere on the site, and no trial or self-serve signup path (https://turbomcp.ai/contact, https://turbomcp.ai/book-a-demo). /pricing, /docs, and /security all return HTTP 404. This is a normal shape for enterprise infrastructure software sold by demo, but it means a buyer cannot scope cost or read technical documentation before talking to sales.
The public storefront looks unfinished
Turbo MCP's own changelog page (https://turbomcp.ai/changelog) lists four version entries dated September–December 2024, before turbomcp.ai was even registered (the domain's registration date is 23 July 2025, confirmed via RDAP at rdap.identitydigital.services). The listed features — task templates, mobile app optimization, an analytics dashboard with "interactive charts" — don't describe an MCP gateway at all; this reads as unedited placeholder copy left over from the page's Framer template, not a real release history. The GitHub organization "turbomcp" exists but has zero public repositories (https://github.com/turbomcp), and we could not find any company name, founder, or press coverage tied to Turbo MCP independent of the marketing site itself. Whether Turbo MCP is a Dylibso rebrand or a separate party that took over the domain is not established — Dylibso's own site (dylibso.ai) is now a single-page logo splash with no product links or mention of mcp.run or Turbo MCP.
What happened to the original mcp.run registry
Dylibso's own GitHub organization confirms the original product's fate. dylibso/mcp.run-servlets — described as "Official servlets for mcp.run published by @dylibso" — is now archived (GitHub API returns "archived": true; last push 20 November 2025, no archive date is exposed by GitHub's API) (https://api.github.com/repos/dylibso/mcp.run-servlets). A related client library, dylibso/mcpx4j ("Java client library for https://mcp.run"), is not archived and was last pushed 31 October 2025, but its listed homepage — https://docs.mcp.run/ — no longer resolves at all (https://api.github.com/repos/dylibso/mcpx4j). Taken together, the registry, its documentation, and its official tool library all stopped moving within weeks of each other in late 2025, consistent with the domain being repointed to Turbo MCP shortly after.
Frequently asked questions
What is mcp.run?
Mcp.run is an enterprise AI connectivity platform for IT teams that runs a standards-compliant MCP gateway and management layer between apps and AI tools. It combines The Whole Org Chart, Audit & Control, and Advanced Security to govern access across internal systems. The platform is self-hosted and cloud-ready, integrates with OIDC-compatible identity providers, and includes Continuous Updates, Audit Usage, and 1-click Deactivation.
What is mcp.run used for? Who is it for?
Mcp.run is used for The Whole Org Chart, Audit & Control, and Advanced Security. It's built for Enterprise IT teams, Security and compliance teams, and Platform engineers.
Editor's read
Check how much of your identity stack is already OIDC-compatible before rollout. Mcp.run is built around that integration, so confirming your provider and access policies line up will determine how smoothly centralized control and deactivation work across the organization.
