Today, Vantage is open‑sourcing its Remote Model Context Protocol (MCP) Server, so teams can connect AI agents to live cloud cost and usage data through a standard, secure interface. Customers can now review and extend the server, run it in their own environments, and combine it with other MCPs to automate FinOps workflows.

Before today, customers had to choose between running a local, open‑source MCP with full visibility into the code and tools but requiring local installation, or using Vantage's hosted Remote MCP without transparency into the underlying implementation. These options worked well for quick adoption, but they limited flexibility for teams that wanted both the convenience of remote deployment and the ability to inspect, audit, or customize the server code.

Now, with an open‑source Remote MCP, customers retain the convenience of the Remote MCP feature set while gaining full control. They can self‑host locally via stdio for direct communication with MCP clients, or deploy remotely as a Cloudflare Worker with HTTP endpoints supporting multiple authentication methods. Customers can inspect and contribute to the code, adapt tools to match their FinOps practices, and choose the deployment mode that best fits their needs—whether for individual development, team collaboration, or production environments, all while using familiar MCP clients like Claude Desktop and Cursor.

The open-source Remote MCP is available today. To get started, follow the repository's README to run locally via stdio or deploy remotely using multiple authentication methods.

Frequently Asked Questions

1. What is being launched today?

Vantage is open‑sourcing the Remote MCP Server, providing the same tools as our hosted Remote MCP with the ability to self‑host and extend.

2. Who is the customer?

FinOps and engineering teams that want agentic access to cost and usage data, with the flexibility to run, audit, and customize the server.

3. How much does this cost?

There is no additional cost for using the Remote MCP Server. Standard Vantage API rate limits apply when querying your data.

4. How is this different from the hosted Remote MCP?

This update is functionally similar in read‑only scope. The Hosted Remote MCP provides managed OAuth, secure transport, and zero‑maintenance upgrades. The open‑source server gives you code transparency and control to self‑host and customize.

Note: The previous version of the vantage-mcp-server repo code is available here.

5. What functionality is included?

The MCP has tools to query and list costs and cost reports, get forecasts, budgets, dashboards, tags, tag values, anomalies, unit costs, and more. It can also create Cost Reports and budgets. Additional tools are consistently being added, so be sure to check the repository for an up-to-date list of tools.

6. Which MCP clients are supported?

Any client that implements MCP, including Claude, Cursor, Bedrock, Goose, and others. See each client’s version requirements for details on how to connect.

7. How do I authenticate?

Use a Vantage MCP Server API token for self‑hosted deployments, which can be provisioned in Vantage Settings. The hosted Remote MCP uses OAuth and scopes permissions per user.

8. Does Vantage see my prompts?

Prompts remain within your AI agent environment. Vantage only observes API calls made on your behalf when you query your own data.

9. Can I contribute?

Yes—open issues and pull requests in the repository. We welcome bug reports, docs updates, and proposals for new tools.

10. How do I get started?

Clone the repository, configure your API token, and connect from an MCP client, like Claude Desktop or Cursor. See the README for a quickstart and deployment steps.