Everstake MCP Server

MCP serverAI & models

Once this app is added, your AI can look up staking data across more than 130 networks. It works without holding or moving any of your assets, since the data is non-custodial. Just ask your AI about staking on a network and it can find the details for you.

Available today. Use it from your connected AI after setup.

Add the app, then ask your AI a staking question about any network it covers. Start with a network you already follow to see what data it can pull up.

Then ask your AI: use Everstake MCP Server to get chains

What your AI can do with it

  • Look up staking data across 130+ networks
  • Answer staking questions about a specific network
  • Pull staking details for several networks in one conversation
  • Check staking information without your assets ever being held or moved

From the project's README

As published by everstake/mcp in README.md.

MCP server exposing Everstake staking data and company information to AI agents. Built in Go using modelcontextprotocol/go-sdk.

Transports: Streamable HTTP (MCP 2025-03-26 spec, single / endpoint) or stdio. Selected via MCP_TRANSPORT.


Available Tools

ToolTypeDescription
get_company_profilestaticCompany overview, metrics, certifications
get_productsstaticProduct details: Institutional Staking, VaaS, Yield, SWQOS, ShredStream
get_solutionsstaticSolutions by audience: custodians, exchanges, asset managers, banks, fintech
get_developer_docsstaticSDK links, integration guides, API references
get_contact_informationstaticContact channels and routing guide
get_security_profilestaticCertifications: SOC 2 Type II, ISO 27001, NIST CSF, ITGC, GDPR, CCPA
get_integrationsstaticCustody integrations: Fireblocks, BitGo, Anchorage, Coinbase, etc.
get_uptime_metricsliveUptime metrics from dashboard API (30 min cache)
get_chainsliveSupported chains with APY, fees, status (30 min cache)
staking_calculatorliveEstimated staking rewards by network and amount
request_integrationwriteSubmit integration/staking inquiry to Everstake sales

Running the Server

Prerequisites

  • Go 1.26.1+
  • Environment variable DASHBOARD_URL set (required)

Local

export DASHBOARD_URL=https://your-dashboard-api-host
go run ./cmd/mcp_server

The server starts on port 8080 by default. Override with PORT=<port>.

Stdio mode

export DASHBOARD_URL=https://your-dashboard-api-host
export MCP_TRANSPORT=stdio
go run ./cmd/mcp_server

In stdio mode, the HTTP server, /health endpoint, and rate limiting are disabled. Logs go to stderr; stdout carries the MCP JSON-RPC protocol.

Docker

HTTP mode:

docker build -t everstake-mcp .
docker run -e DASHBOARD_URL=https://your-dashboard-api-host -p 8080:8080 everstake-mcp

Stdio mode (interactive, no port mapping):

docker run --rm -i \
  -e DASHBOARD_URL=https://your-dashboard-api-host \
  -e MCP_TRANSPORT=stdio \
  everstake-mcp

Use -i to keep stdin attached for the JSON-RPC protocol; do not pass -t (a TTY breaks line-based JSON framing). For MCP clients that launch the server as a subprocess, point command at docker with args matching the above, or wrap in a helper script (see .vscode/stdio_docker.sh for an example using --env-file).

Environment Variables

VariableDefaultRequired
DASHBOARD_URLyes
MCP_TRANSPORThttpno (http or stdio)
PORT8080no (http mode only)
GIN_MODEno (release set in Dockerfile)

Health Check

GET /health

Debugging with MCP Inspector

MCP Inspector is the official tool for poking at MCP servers — list tools, call them, see raw JSON-RPC.

Stdio note: stdio servers are owned by the parent process; you cannot connect Inspector to an already-running stdio container. Inspector must launch the subprocess itself. For an interactive session you can reconnect to, use HTTP mode (option below).

HTTP mode (recommended for general debugging)

Start the container:

docker run --rm -p 8080:8080 \
  -e DASHBOARD_URL=https://your-dashboard-api-host \
  -e DASHBOARD_API_KEY=<key> \
  everstake-mcp

Open Inspector:

npx @modelcontextprotocol/inspector

In the UI:

  • Transport Type: Streamable HTTP
  • URL: http://localhost:8080/

Container stays running across Inspector sessions.

Stdio mode (verify stdio transport)

Let Inspector launch the container so it owns stdin/stdout:

npx @modelcontextprotocol/inspector \
  docker run --rm -i \
    --env-file .env \
    everstake-mcp

Or fill the UI manually:

  • Transport Type: STDIO
  • Command: docker
  • Arguments: run --rm -i --env-file .env everstake-mcp

Each Inspector session forks a fresh container; quitting Inspector kills it.


Linting

The project uses golangci-lint with a strict configuration in .golangci.yml.

Install golangci-lint:

go install github.com/golangci/golangci-lint/cmd/golangci-lint@latest

Run:

make lint

Key rules enforced: staticcheck (all checks), gosec, gocritic (diagnostic/style/performance tags), revive (40+ rules including early-return, error-strings, var-naming), errchkjson, bodyclose, contextcheck, and more. nolintlint requires specific lint directives — bare //nolint is not allowed.


Editing Tool Responses

Static tools

Edit tools.yaml. Each map key is the tool name; the static_response field is returned verbatim to the AI agent.

tools:
  get_company_profile:
    description: |
      ...
    static_response: |
      COMPANY: Everstake
      ...

To add a new static tool:

  1. Add an entry under tools: in tools.yaml with static_response.
  2. Add a corresponding field to ToolsConfig in internal/config/mcp_config.go with a matching yaml struct tag — the name is injected automatically via reflection.
  3. Register it in internal/server/mcp/server.go using staticTextTool().

Content rules

Cross-cutting rules that apply to all tool responses are in .vscode/tools_src/RULES.md. These cover:

  • Certification differentiator language
  • Non-custodial positioning
  • Vault product disclaimers
  • APY disclaimer wording
  • Lead source tagging for request_integration

Dynamic tools

get_uptime_metrics and get_chains fetch live data from the dashboard API with a 30-minute in-memory cache. Their handlers are in internal/server/mcp/dashboard.go. The underlying API client lives in pkg/everstake/dashboard/.

Tools it offers (11)

What this server listed when ahel dialed its public endpoint in Sep 2026, with no key and no account of yours. The names are the server’s own.

  • get_chains
  • get_company_profile
  • get_contact_information
  • get_developer_docs
  • get_integrations
  • get_products
  • get_security_profile
  • get_solutions
  • get_uptime_metrics
  • request_integration
  • staking_calculator

Signals

GitHub stars
1
Last commit
Aug 2026
Advanced
Delivery
mcp MCP server → your ahel gateway (mcp.ahel.ai) → every connected AI client.
Catalog kind
mcp-server
Gateway key
one-everstake-mcp
Source
github.com/everstake/mcp
Hosted endpoint
https://mcp.everstake.one