@savvly/mcp-server
v1.0.109
Published
Explore Savvly MCP: retirement projections, eligibility, comparisons, and FAQs for an SEC-registered fund that pairs full market upside with later-life longevity payouts.
Readme
@savvly/mcp-server (npm)
Explore Savvly MCP: retirement projections, eligibility, comparisons, and FAQs for an SEC-registered fund that pairs full market upside with later-life longevity payouts.
Savvly offers longevity-linked, SEC-registered investment products — built on the Savvly 80+ fund — that invests pooled contributions in a low-cost S&P 500 ETF and makes milestone cash payouts at ages 80, 85, 90, and 95, turning a longer life into a financial reward rather than a risk.
Savvly MCP lets financial advisors, planning tools, and AI agents query fund data, model projections, and compare against alternative retirement products directly, in real time, in any conversation, and without any manual lookup.
This is the npm distribution of the same server also published to PyPI as
savvly-mcp-server, to NuGet as Savvly.McpServer, and to the official MCP
Registry as com.savvly/savvly.
Install & run
npx @savvly/mcp-server
# or install globally:
npm install -g @savvly/mcp-server && savvly-mcpClaude Desktop / Cursor config
{
"mcpServers": {
"savvly": {
"command": "npx",
"args": ["@savvly/mcp-server"]
}
}
}Tools
| Tool | Description |
|------|-------------|
| get_savvly_product_info | Complete product details |
| compare_savvly_vs_alternative | Comparison vs annuities / target-date funds / etc. |
| project_savvly_lumpsum | Lump-sum payout projection |
| project_savvly_monthly | Monthly contribution projection |
| project_retirement_with_savvly | Full retirement simulation |
| check_savvly_eligibility | Eligibility check |
| get_savvly_faq | FAQ filtered by topic |
| search_savvly_content | Audience-tagged Q&A library search |
Projection tools emit interactive MCP Apps chart widgets on hosts that support them.
Resources
savvly://product/overviewsavvly://product/comparison-matrixsavvly://product/payout-schedulesavvly://content/qa-library
Usage records
This package keeps no log of its own. It sends Savvly a usage record for each
tool call, in small batches, to https://api.savvly.com/usage/v1. A record
holds only the tool's name, whether the call succeeded (and for a failure a
fixed reason code such as estimator_unreachable), how long it took in
milliseconds, the name your MCP host reports for itself (such as claude-ai),
and this package's language and version. It never holds the tool's arguments,
a hash of them, the result, or an error's text. Sending happens in the
background, never slows or fails a tool call, and drops a record it cannot
send.
To turn it off, set SAVVLY_TELEMETRY=0 (or DO_NOT_TRACK=1) in this
server's own env settings in your MCP host:
- Claude Desktop with the
.mcpbextension: switch off Send usage records in the extension's settings. - Hosts with a JSON config (Claude Desktop's
claude_desktop_config.json, Cursor's~/.cursor/mcp.jsonand others): add"env": { "SAVVLY_TELEMETRY": "0" }beside"command"in the server's entry. - VS Code: the same
envkey in the server's entry inmcp.json. - Claude Code:
claude mcp add savvly -e SAVVLY_TELEMETRY=0 -- npx -y @savvly/mcp-server. - Docker: pass it to the container,
docker run -i --rm -e SAVVLY_TELEMETRY=0 ghcr.io/savvly/savvly-mcp-server.
A variable set in your shell often does not reach the server: MCP hosts start a
stdio server with only a few variables (such as PATH and HOME) plus the
ones in its own env settings, and an app started from the Dock or Start menu
never sees your shell's variables. So set it in the server's env settings,
which always reach it, not in your shell.
Remote alternative
Prefer a hosted endpoint? Connect via Streamable HTTP at https://api.savvly.com/mcp.
License
Apache-2.0.
