Everstake Launches Open-Source MCP Server: Staking Data, Now Native to AI Agents

TL;DR

• The new Everstake MCP server lets any MCP-compatible AI assistant query staking data, calculate rewards, check chain status, and submit integration inquiries through natural language. 
• Built in Go using the official modelcontextprotocol/go-sdk, it implements the MCP 2025-03-26 spec.
• 11 tools across 3 categories (live data, knowledge, action).
• Real-time data with a 30-minute cache for performance.
• Covers 130+ networks and a 99.98% uptime track record.
• Backed by SOC 2 Type II, ISO 27001:2022, NIST CSF, ITGC, GDPR, and CCPA compliance.
• Free, open-source, and production-ready.

Everstake has released an open-source Model Context Protocol (MCP) server that gives AI agents direct access to live staking infrastructure data. 

The server is live at https://mcp.everstake.one and the full source code sits at https://github.com/everstake/mcp, learn more about MCP at https://everstake.one/mcp 

What Is an MCP Server?

The Model Context Protocol is an open standard that lets AI assistants connect to external tools and data sources in a structured, secure way. It works as the universal API for AI systems.

Instead of writing a custom integration every time a new tool needs to talk to a language model, developers connect once via MCP, and the agent handles the rest. The protocol was designed to make AI use predictable, auditable, and portable across models like Claude, GPT, and others.

An MCP server, in turn, is the piece of infrastructure that exposes a specific set of tools and data to those agents. The Everstake MCP server shares staking-related data.

Reason Behind MCP Server for Staking

Staking data is fragmented, updating in real time, and operationally sensitive, which makes it a poor fit for static documentation. When an AI agent needs current chain status, rate information, or uptime metrics, scraping a website or hitting an undocumented endpoint is brittle.

Everstake built and open-sourced this server so that AI agents can pull live staking data, calculate rewards, check chain status, and submit integration inquiries through one structured interface.

For institutions, this also changes due diligence. A compliance team or partnership lead can ask an AI assistant about Everstake certifications, custody integrations, or uptime, and the answer comes from a verified source rather than a model’s training data.

What the Server Offers

The Everstake MCP server exposes 11 tools grouped into 3 categories: live data, knowledge, and action. Each tool is purpose-built for a specific kind of query an AI agent or developer might run.

The breakdown below covers every tool currently available.

Live Data Tools

Live tools query the Everstake dashboard API in real time, with a 30-minute in-memory cache to keep performance fast and load predictable.

Tool	What It Returns
get_chains	All supported chains across 130+ networks, with APR data, fee structures, and status.
get_uptime_metrics	Real-time uptime data pulled directly from Everstake monitoring infrastructure (99.98% record).
staking_calculator	Estimated rewards for any supported network and amount.

Static Tools

Static tools return static, always-accurate content covering what an AI agent or developer needs to know about Everstake.

Tool	What It Covers
get_company_profile	Company overview, headline metrics, and credentials.
get_products	Institutional Staking, Validator-as-a-Service, Yield Infrastructure products, SWQOS, ShredStream.
get_solutions	Context tailored for custodians, exchanges, asset managers, banks, and fintech teams.
get_security_profile	Full compliance stack: SOC 2 Type II, ISO 27001:2022, NIST CSF, ITGC, GDPR, CCPA.
get_integrations	Custody integrations including Fireblocks, BitGo, Anchorage, and Coinbase Prime.
get_developer_docs	SDK links, API references, and integration guides.
get_contact_information	Routing for sales, support, and partnership inquiries.

Write Tools

Write tools let an agent do something on behalf of a user, not just retrieve data. There is currently one tool in this category.

request_integration submits a staking or integration inquiry directly to the Everstake sales team on behalf of a user.

How It Works Under the Hood

The server is built in Go using the official modelcontextprotocol/go-sdk and implements the MCP 2025-03-26 spec with a single Streamable HTTP endpoint at /. 

The architecture is intentionally lean so that the server is auditable, easy to self-host, and quick to extend.

The request flow follows four steps:

1. An AI agent connects to the MCP server.
2. The agent invokes one of the 11 available tools.
3. The server either returns a predefined static response or fetches data from the Everstake dashboard API.
4. The agent uses the response to generate output or trigger an action.

Types of Requests an Agent Can Make

The tool set is designed so that almost any reasonable staking question maps cleanly onto one or two calls. Below is a snapshot of how natural-language queries route to MCP tools.

User Question	Tool Invoked
“What’s the current rate on Ethereum?”	get_chains
“How reliable is Everstake infrastructure?”	get_uptime_metrics
“Estimate rewards if I stake 100 SOL for a year.”	staking_calculator
“Which custody platforms integrate with Everstake?”	get_integrations
“Is Everstake SOC 2 certified?”	get_security_profile
“Connect my team with Everstake sales.”	request_integration

Agents can answer due-diligence and operational questions without falling back on outdated training data or generic web search.

Who It’s For

The server is built for three primary audiences. The needs of developers, institutions, and agent builders differ, but the same 11 tools serve them all.

• Developers building AI-powered applications that need staking data without writing bespoke clients. Plug in the MCP server and the agent can answer questions about rates, uptime, and rewards.
• Institutions evaluating Everstake can use any MCP-compatible assistant to perform due diligence directly against verified sources, instead of hunting through documentation.
• AI agent builders who want staking infrastructure as a first-class capability inside their tech.

Everstake: A Reliable Validator Optimizing for an Agentic Future

Everstake is one of the world’s largest non-custodial staking providers, having supported over $7B in staked assets across 130+ networks to date with 99.98% uptime and 1.6M historical delegators. The infrastructure is certified/attested against SOC 2 Type II, ISO 27001:2022, NIST CSF, ITGC, GDPR, and CCPA standards.

The MCP server is one of the first specific steps Everstake has taken to operate in an agentic future, where AI systems handle the queries, integrations, and workflows that humans run today. 

Try It Now

Any client implementing the MCP 2025-03-26 specification, including but not limited to Claude Desktop, ChatGPT, Claude Code, Cursor, VS Code, Windsurf, Continue, Cline, Goose, and Zed. The server uses Streamable HTTP transport and exposes a single endpoint.

The server is live at https://mcp.everstake.one, and the full source code is on GitHub at https://github.com/everstake/mcp.
To learn more about MCP, visit https://everstake.one/mcp.

FAQ

Does it require an API key?

No. Read tools like live data on chains, APR, uptime, and the staking calculator are publicly accessible without authentication.

What data does the server return?

Live staking data (chains, APR, fees, validator status, uptime metrics), staking reward calculations for any supported network and stake amount, company and product information, security and compliance profile, custody integrations, developer documentation references, and contact routing. 

Is it non-custodial?

Yes. Everstake operates on a fully non-custodial basis across all supported networks. Your organisation retains control of keys and assets at all times, and Everstake only operates the validator infrastructure. The MCP server itself does not move funds, sign transactions, or hold keys.

Which AI clients does Everstake MCP support?

Any client implementing the MCP 2025-03-26 specification.

What is the uptime SLA?

Everstake maintains a 99.98% observed uptime across all major networks, backed by geo-distributed bare-metal infrastructure with automated failover and continuous monitoring. The MCP server itself surfaces live uptime data through the get_uptime_metrics tool. Custom SLAs are available for institutional partners through Validator-as-a-Service agreements.

How to get enterprise support?

Institutional partners receive engineering coverage, dedicated dashboards, audit-ready reports, and account managers experienced in institutional frameworks. To discuss enterprise support, custom integration, or Validator-as-a-Service deployment, contact our team at everstake.one/contact-us or use the request_integration tool directly from your AI client.