mcp-fsolar
MCP server that connects to Felicity Solar cloud battery data, allowing natural language queries about battery status, cell voltages, and charge state.
README
mcp-fsolar
MCP server + REST API + JS client for Felicity Solar cloud battery data (shine-api.felicitysolar.com).
Ask Claude things like "what's the battery SOC?", "is any cell imbalanced?", or "show me the cell voltages for Bat2" — live from the Felicity cloud.
Install
npm install -g fsolar-mcp
Then create a .env file with your credentials (see Configuration) and start the server:
fsolar-mcp
Or run without installing:
npx fsolar-mcp
Setup from source
git clone https://github.com/RicardoSantos/mcp-fsolar
cd mcp-fsolar
npm install
cp .env.example .env # fill in your credentials
node server.js
Configuration
.env variables:
| Variable | Required | Default | Description |
|---|---|---|---|
FELICITY_USER |
Yes | — | Felicity Solar account email |
FELICITY_PASS |
Yes | — | Felicity Solar account password |
FELICITY_PORT |
No | 3010 |
HTTP server port (standalone) |
FELICITY_POLL_MS |
No | 30000 |
API poll interval in ms (standalone) |
FELICITY_SNAPSHOT_ENABLED |
No | true |
Enable background snapshot poller |
FELICITY_SNAPSHOT_MS |
No | 600000 |
Snapshot interval in ms (min 60 000, max 3 600 000) |
FELICITY_SNAPSHOT_DAYS |
No | 3 |
Days of intra-day snapshots to retain (min 1, max 30) |
FELICITY_DAILY_DAYS |
No | 90 |
Days of daily snapshots to retain (min 7, max 365) |
Algorithms & metrics
Detailed documentation of every derived metric (formulas, thresholds, assumptions, webhook events) is in ALGORITHMS.md.
Wire into Claude Code
After starting the server, register it:
claude mcp add felicity --transport sse http://localhost:3010/sse
Or add to .claude/settings.json:
{
"mcpServers": {
"felicity": {
"type": "sse",
"url": "http://localhost:3010/sse"
}
}
}
To auto-start with Claude Code (no separate process needed):
{
"mcpServers": {
"felicity": {
"command": "npx",
"args": ["fsolar-mcp"],
"env": {
"FELICITY_USER": "your@email.com",
"FELICITY_PASS": "yourpassword"
}
}
}
}
MCP tools
| Tool | Description |
|---|---|
get_all_batteries |
Live status of all batteries — SOC, power, voltage, temperature, charging state |
get_battery |
Detailed status of one battery by alias (Bat1/Bat2/Bat3) or serial number |
get_cell_voltages |
Individual cell voltages (mV) — useful for detecting cell imbalance |
get_fleet_summary |
Compact health summary: total energy, worst cell delta, temperatures |
get_balance_trend |
Cell delta trend over the last ~60 min (improving / stable / degrading) |
get_snapshots |
Raw snapshots for the last ~60 min (one per ~10 min) |
REST API
The server also exposes a plain HTTP API on the same port:
GET /batteries # all batteries
GET /batteries/:id # one battery by alias or serial number
GET /sse # MCP SSE endpoint
Pass X-Last-Fetched-At: <ISO timestamp> to skip the cache when you already hold fresh data.
JS client library
import { FelicityClient, MemoryCacheAdapter } from './index.js'
const client = new FelicityClient({
user: 'you@example.com',
pass: 'yourpassword',
cache: new MemoryCacheAdapter(),
ttl: 30, // seconds
})
const { batteries } = await client.getBatteries()
const { battery } = await client.getBattery('Bat1')
Full TypeScript types are in index.d.ts.
Probe (debug)
Dumps raw API responses for every device — useful for exploring your setup:
node probe.js
How it works
The Felicity cloud API requires passwords to be RSA-encrypted (public key extracted from the Android APK). The client handles login, token refresh, and caching automatically. A background poller keeps data fresh so MCP tool calls are instant.
Available data
Per battery
| Field | Type | Description |
|---|---|---|
sn |
string | Serial number |
alias |
string | Human name (Bat1, Bat2, Bat3) |
model |
string | Model string from BMS |
status |
NM | AL | FL | OF |
Normal / Alarm / Fault / Offline |
soc |
number % | State of charge |
soh |
number % | State of health |
voltage |
number V | Pack voltage |
current |
number A | Pack current |
power |
number W | Pack power (positive = charging, negative = discharging) |
chargingState |
string | charging / discharging / standby |
remainingKwh |
number | Estimated remaining energy |
capacityAh |
number | Rated capacity (Ah) |
ratedEnergyKwh |
number | null | Rated energy (kWh) |
isBalancing |
boolean | Active cell balancing in progress |
warningCount |
number | Active BMS warnings |
Cell & temperature data
| Field | Type | Description |
|---|---|---|
cellVoltages |
number[] mV | All 16 cell voltages |
cellVoltageMin |
number | null mV | Lowest cell voltage |
cellVoltageMax |
number | null mV | Highest cell voltage |
cellDelta |
number | null mV | Spread between min and max cell (imbalance indicator) |
minCellNum |
number | null | 1-based index of weakest cell |
maxCellNum |
number | null | 1-based index of strongest cell |
cellTemps |
number[] °C | 4 physical temperature sensors |
tempMin |
number °C | Lowest sensor reading |
tempMax |
number °C | Highest sensor reading |
BMS protection limits
| Field | Type | Description |
|---|---|---|
chargeVoltLimit |
number | null V | Max charge voltage |
dischargeVoltLimit |
number | null V | Min discharge voltage |
chargeCurrLimit |
number | null A | Max charge current |
dischargeCurrLimit |
number | null A | Max discharge current |
BMS lifecycle counters
| Field | Type | Description |
|---|---|---|
batCycleIndex |
number | null | Total charge cycles |
batFullCount |
number | null | Times battery reached full charge |
batUnderVoltageCount |
number | null | Under-voltage events |
Module breakdown (per battery)
Each battery exposes up to 4 modules, each containing:
| Field | Type | Description |
|---|---|---|
index |
number | Module number (1–4) |
cells |
number[] mV | 4 cell voltages in this module |
temp |
number | null °C | Physical sensor for this module |
min / max / delta |
number mV | Voltage spread within the module |
Balance trend (historical)
Computed from snapshots taken every ~10 min (up to 12 h retained):
| Field | Description |
|---|---|
direction |
improving / stable / degrading |
deltaChange |
mV change newest − oldest (negative = improving) |
history |
cellDelta values oldest → newest |
balancingCount |
Snapshots where balancing was active |
currentBalancingStreak |
Consecutive trailing snapshots with balancing on |
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。