We're constantly updating Metorial. Here's what's new.
Oracle Fusion Cloud, Granola, Google Chat, Google Compute Engine, MotherDuck, Companies House, Handelsregister, Openregister, Destatis and KSeF join the catalog. Slack, GitHub, Google, Microsoft, Dynamics 365, Stripe, Xero, QuickBooks and other existing integrations also got updates. The catalog now has 1,100+ MCP servers and 20,000+ tools.
Sandboxes give an agent its own isolated Linux machine, with its own CPU, memory and disk. Start from a managed image such as Debian, Node.js or Python, or build your own from a Dockerfile, then run commands and move files from the CLI or the TypeScript and Python SDKs. Disks can outlive a machine, machines can pause or hibernate when idle, and ports are served over HTTPS with access tokens or open to the public. Outbound firewalls control what a machine can reach. Read the Sandboxes docs to get started.
When you grant access to an integration, you can now choose tools by category instead of one by one, and see allowed tools grouped by category. Access for an integration is granted per app, so the same rules apply wherever the agent connects.
Chat integrations connect an agent to a chat tool, starting with Slack, so people can work with it in a conversation. A new Chat section in the dashboard shows and manages them.
Integrations can now send events to Metorial, starting with Slack. Route them to event destinations and webhooks under the Developer section so an agent or your own system can react to what happens instead of polling.
Tools can now return and accept file attachments, with stable, checked links so an agent can hand a file back without the link breaking later.
An Outpost is a small open-source proxy (Apache 2.0) that you run inside your own network. It connects out to Metorial, so internal systems are reachable without opening inbound access. Add middleware to redact, block or log MCP traffic on the way through.
Audit logs now cover portals, skills and Workforce, include IP information, and have retention settings by plan. Audit log streams send events to your own systems.
Decide per project whether Metorial keeps full message data, intent only, or none at all.
Custom encryption keys are now available as a plan add-on, building on June's KMS support. Support access can be granted to Metorial for a limited time, extended or revoked, with a history of every grant.
Serve Portals from your own domain, add a marketplace inside a portal with a manager UI, and set a custom Magic MCP URL.
Choose which SSO connections, groups and roles grant access to a workspace.
Set up SSO per connection with SCIM directories for users, groups and roles, with sync logs and a test sign-in so you can check a setup before rolling it out. Part of Metorial Workforce.
Edit a skill together in real time, propose and review changes with merge requests, and import skills from Git: GitHub Enterprise, GitLab Self-Managed, Bitbucket Cloud and Bitbucket Data Center. Skills are now available to every organization.
The employee portal has new Agents, Connections and Operations views with split-view lists and an account page. Each AI client gets its own Magic MCP setup instructions.
Organize people into workspaces and groups, and manage access to each in Metorial Workforce.
You can now import your own AWS KMS keys to encrypt the credentials stored in Metorial Vault. By default Vault manages encryption for you; with KMS you take ownership of the keys, so secrets are encrypted under keys only you control. Every operation involving a KMS-protected secret is logged and auditable.
You can now control how tool names are generated for providers you connect through Magic MCP. The default uses a combination of the provider name and the tool's key, which keeps names predictable and consistent across integrations. If that doesn't fit your setup, you can define custom names instead. The setting can be controlled per integration or globally.
Agent Explorer is a new way to test integrations alongside the existing Explorer. Where Explorer lets you browse an integration and call tools yourself, Agent Explorer hands the integration to a real AI model and shows you what it actually does with it. You can run the same integration across Claude, ChatGPT, Gemini, Grok, and others to see how each model behaves before anything goes to production. Every test run goes through the same access control and tracing as live traffic.
Metorial Portals no longer requires an Enterprise plan. Teams on Metorial Scale can now deploy Portals and give employees governed access to company integrations through a single Magic MCP URL. Enterprise keeps everything it had; Scale now gets the same portal setup, access groups, and policies.
Integrations are now a first-class concept on Metorial. Each integration brings together authentication, configuration, credentials, and identity in one place. Making it a single source of truth whether an agent reaches a provider through the API, through Metorial Portals, or through Magic MCP. Start from 1000+ verified integrations or add your own internal tools and private APIs, and govern access with the groups and policies you already manage.
Skills are now available on Metorial. Teams can build reusable agent capabilities, work on them together, and share them, either privately within their organization or publicly through the marketplace. Skills plug into Metorial Portals to extend what agents can do in customer-facing deployments, and into Metorial Workforce to give internal agents a shared, versioned library of capabilities they can draw on.
Metorial Portals gives every employee a single place to find the integrations their company has approved and connect them to any agent through a single Magic MCP URL. Sign in with company SSO, no tokens to manage. Admins control who sees what through access groups and policies, and every interaction is fully traced. Portals is part of Metorial Workforce, so the integrations and access rules you configure there are what your employees' agents work with.
A dedicated EU region for teams that need compute and data closer to European users, with a clearer data residency story. Granular access and roles; files as part of workflows; scale, security, and hardened auth; a catalog that syncs reliably and spotlights strong integrations.
V2 brings new documentation, SDKs, and APIs built for today; V1 is fully behind us on customer and marketing paths. Tighter billing and membership; dashboards, webhooks, and secure handoff for operators; a public integration registry so teams can discover integrations without hunting through side channels.
Environments, custom servers, containers, observability, and faster setup paths for integrations in production. Node and Python SDKs advanced; documentation moved over in bulk; the public site and integration pack generation are easier to run and maintain.
A new in app landing matches the web story so evaluation and daily use feel like one product. Partner pages (Meteor Computing, Protoguard) and a docs platform replace scattered pages with one place to look.
In addition to the existing TypeScript or JavaScript based managed MCP servers, Metorial now supports Python-based ones. This one is for the Python lovers :)
Businesses and enterprises can now connect their SAML or OAuth based authentication services, like Okta, Entra, and Rippling, to Metorial. Metorial's authentication infra then passes along the authentication state to our MCP platform to enable true single sign on for MCP.
Until now Metorial was focused on API-controlled access to MCP servers. Today we're invading Cursor, Claude Code, and more with Magic MCP. Magic MCP is a single URL you can use to connect all of your agents and vibe coding tools to our reliable platform.
MCP servers don't support webhooks from sources like Stripe or Github. From now on, they do using Metorial Callbacks. Callbacks allow MCP servers to receive webhooks from external providers and the forward them to your agents to process them. It's as easy as one line of code and we handle the rest, including monitoring, retires, and more.
Yup, we just open sourced Metorial. Read the code, contribute, or get inspired. Check it out at github.com/metorial/metorial-platform
Starting today, anyone can publish MCP servers on Metorial. Rest assured that these servers are vetted and verified by the Metorial team. This now enabled teams building their own MCP servers on Metorial to share them with the community. We are already working with select MCP server providers to publish their servers on our platform.
Enterprise deployments need advanced security for MCP servers. Metorial already isolates every MCP connection, connecting it to it's own MCP server isolated from others to avoid problems like token leakage. Today we are taking this one step further by allowing you to restrict where an MCP server can connect to so you can be assured it only connects where it is supposed to.
With managed servers you can now deploy your own MCP servers on Metorial. These servers run on the same super reliable platform that already powers our built-in MCP servers. Get started by creating a new managed server in the Metorial dashboard - it deploys in seconds and scales to thousands of connections with monitoring and our great SDKs on top.
With oauth connections you can now set up authentication to MCP servers on Metorial using OAuth. This works with any MCP server - not matter if it natively supports OAuth or now. We handle all of the complicated parts like authentication errors, token refresh, and security.
Starting today, every MCP server running on Metorial is running on our new custom MCP engine. Think of this as our own Kubernetes built just for MCP. It's the scaling and orchestration layer for serverless MCP. What this means for you is that MCP connections to Metorial are now more reliable, faster, and you can have tens of thousands of them.
Providers like Linear or Neon offer their own remote MCP servers. Metorial now has the ability to connect these providers just like you would connect any other MCP server and you get our SDKs, monitoring, and reliable platform on top.
Metorial's new Python and TypeScript SDKs enable you to connect the Metorial Integration Platform to the AI Agent SDKs you are already using. We have support for Vercel's AI SDK, the OpenAI, Anthropic, and Google SDKs, and even LangChain.
Check out our examples on GitHub:
Connect to any integration hosted on Metorial using our serverless gateway. When you connect, we boot up an MCP server in less than a second. Connections can run for day or even months while only incurring compute costs when they are needed. You can even reconnect to MCP connections weeks of months after they have been started. Of course all of this is integrated with Metorial APIs and SDKs to you can connect from JS or Python in just one line of code.
The Metorial Integration Platform makes it easy to connect any AI agents to external tools and data sources through MCP. All of this is enabled by our serverless platform specifically designed for MCP. The Metorial platform can run any MCP server with sub-second cold starts. Things like monitoring, security, and scaling are handled automatically.
We’ve introduced Model Context Protocol (MCP) Containers—a new way to run and manage AI model servers more efficiently. Over 400 MCP servers have now been fully containerized, making it easier to deploy, scale, and maintain AI infrastructure.
The best part: it’s completely open source. You can explore the code, contribute, or start using MCP Containers in your own workflows.
Learn more: https://github.com/metorial/mcp-containers
We’re excited to roll out Chat Message Streaming, bringing faster and more dynamic AI interactions to Metorial Portals and the Metorial API!
This update makes AI conversations feel more natural, fluid, and engaging—whether you’re using Metorial Portals or integrating via the API.
We’ve given the Metorial Dashboard a major upgrade, making it sleeker, faster, and more user-friendly than ever!
🚀 Check out the new dashboard and experience the upgrade today!
We’re thrilled to launch Metorial Portals, a powerful way to create custom ChatGPT-like websites connected to your AI agents—no coding required!
With Portals, you can:
Whether you're building an AI-powered customer support assistant, knowledge base, or interactive chatbot, Metorial Portals make it effortless to deploy.
🚀 Start building your AI-powered website today!
We’re excited to introduce support for image models like DALL·E, enabling powerful AI-driven image generation directly within Metorial!
With this update, you can:
Whether you're building creative tools, automated design assistants, or AI-powered storytelling experiences, Metorial now makes it easier than ever to blend text and image generation for limitless possibilities.
Start creating with AI-powered images today!
We’re excited to launch the Metorial API, giving developers full programmatic control over their AI Agents! Now, you can seamlessly integrate AI-powered interactions into your applications, products, and workflows.
With the Metorial API, you can:
Whether you're building a custom AI-powered app, automating customer support, or integrating AI into existing software, the Metorial API gives you unmatched flexibility and control.
Get started today – Check out our API documentation and start building! 🚀
With the rapid advancements in AI, tools like OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini are more powerful than ever. However, creating complex AI interactions remains a challenge—until now.
Metorial is a high-level visual AI builder that simplifies the process with an intuitive drag-and-drop interface. Whether you're integrating pre-trained models or crafting custom AI agents, Metorial makes AI development accessible to everyone.
Seamless code and no-code integrations ensure that AI-powered tools can be added to any app or website in minutes. Whether you're a developer or a no-code enthusiast, Metorial bridges the gap between APIs and visual AI workflows—empowering innovation without limits.
Be among the first to experience the next generation of AI development. Sign up for updates today!