DEV Community: hardy lee

I made my product database queryable by AI agents — a free API + MCP endpoint

hardy lee — Mon, 10 Aug 2026 02:24:25 +0000

AI assistants are quietly becoming the discovery layer. People don't Google "best power station for a camper van" as much anymore — they ask ChatGPT or Claude. The problem: for a niche like portable power stations, the model usually answers from stale, half-remembered specs and just makes numbers up.

I maintain a structured database of 102 portable power stations (specs, $/Wh, cycle life, a transparent score). So I did the obvious thing: I made it queryable by machines and AI agents — free, no auth.

A plain JSON API (no key, no signup)

# Filter/sort products
curl "https://sunsee.cc/api/products.json?max_price=500&sort=value&limit=5"

# Get one product's score breakdown
curl "https://sunsee.cc/api/score.json?slug=ecoflow-delta-2"

# Size a system from a list of appliances
curl "https://sunsee.cc/api/calculate.json?appliances=[{\"name\":\"Fridge\",\"watts\":60,\"hours\":24}]&region=us-south"

Every response is clean JSON with consistent fields (battery_capacity_wh, battery_type, dollar_per_wh, sunsee_score, …). No HTML scraping, no inconsistent retailer specs.

A discovery endpoint for AI agents

The part I actually care about: a single discovery endpoint that describes the available tools, so an agent can find and call them without me hardcoding anything on its side.

curl "https://sunsee.cc/api/mcp.json"

It returns the tool list (search products, size a system, get a score) with their parameters — the same idea as an MCP tool manifest. Point an agent at it and it can answer "what's the best value station under $500?" or "how big a battery do I need for a fridge + CPAP for 3 days?" with real data.

llms.txt for citation guidance

I also added a llms.txt (like robots.txt, but for language models) telling models when and how to cite the data:

https://sunsee.cc/llms.txt

It maps common questions to the right endpoint, and asks for attribution. No idea yet how much models will respect it — but the standard is young and it costs nothing to publish.

Why bother?

Honestly: if an AI is going to be the thing recommending products, I'd rather it pull transparent, spec-based data (with a public scoring formula, no pay-to-play) than hallucinate. And if it cites the source, that's discovery.

Disclosure: the underlying site has affiliate links — that's how I hope to fund it — but the API, the dataset, the calculator, and the scoring formula are all free and open.

Try it / break it

API playground starts at https://sunsee.cc/api/products.json
Open dataset (CC-BY-4.0): https://github.com/gkin9/portable-power-stations-dataset
The human version (calculator): https://sunsee.cc/calculator

If you build agent/LLM tooling, I'd love to know whether the discovery endpoint is shaped usefully — and what fields you'd want added. Happy to iterate.

LiFePO4 Won the Power Station Wars. Sodium-Ion Wants a Rematch.

hardy lee — Wed, 05 Aug 2026 01:30:34 +0000

When I pulled the specs on 100+ portable power stations recently, one number jumped out: 94% now use LiFePO4 (lithium iron phosphate). A few years ago the market was full of older lithium-ion (NMC). Today LFP has all but won — it's cheaper, safer, and lasts 3,000+ cycles instead of ~500.

But in the comments on that write-up, a good question came up: is this permanent, or is another chemistry going to shake things up? The honest answer: LFP isn't getting dethroned by NMC — that ship has sailed. The chemistry actually worth watching is a newcomer, sodium-ion (Na-ion). And interestingly, it's strongest exactly where LFP is weakest.

Why LiFePO4 won in the first place

Quick recap, because it explains what a challenger has to beat:

Cost — no cobalt or nickel. This is the big one; it's why price-per-Wh has collapsed (median is around $0.61/Wh now).
Cycle life — ~3,000–6,000 cycles vs ~500 for NMC. At daily use that's the difference between ~8 years and ~18 months.
Safety — LFP is far more resistant to thermal runaway, which matters a lot for something sitting in your van or living room.

The one price LFP pays for all that: lower energy density. It's heavier and bulkier per Wh than NMC. That's why featherweight camping units sometimes still use lithium-ion — but for anything that mostly sits still, LFP wins easily.

LiFePO4's real weakness: the cold

There's a second weakness that anyone using one off-grid in winter runs into. LFP does not like the cold:

Usable capacity drops in freezing temps — roughly 10–15% near 0°C, more at deep cold.
More importantly, you must not charge LFP below 0°C / 32°F — doing so causes lithium plating and permanent damage. Good units have a BMS that blocks sub-freezing charging (some add self-heating), but it means your solar might refuse to charge on a frozen morning.

Hold onto that, because it's exactly where the challenger shines.

Enter sodium-ion

Sodium-ion swaps lithium for sodium — an element that's absurdly abundant (it's in seawater and table salt). The pitch for 2026:

Cheaper raw materials — no lithium, cobalt, or nickel, and it can use aluminum current collectors on both sides. As manufacturing scales, the cost floor is potentially below LFP.
Excellent cold performance — this is the headline. Na-ion holds a much larger share of its capacity in freezing temps (some cells keep ~90% at -20°C) and tolerates low-temperature charging far better than LFP. It directly fixes LFP's biggest off-grid weakness.
Safety — very stable thermally, and it can be fully discharged to 0V for safe transport/storage.
Fast charging — generally strong.

Big names are already shipping cells — CATL, BYD's subsidiary, HiNa, and Natron among them — and the first sodium-ion power stations and home batteries have started to appear.

The catch (why LFP isn't dead yet)

Being honest about where Na-ion stands in 2026:

Energy density is still below LFP. For a portable product where weight and size matter, that's a real drawback — a sodium-ion unit is bulkier/heavier for the same Wh. This is the single biggest thing holding it back for portable use.
Cycle life is still catching up. Early cells trailed LFP; newer ones are closing the gap, but LFP's track record is proven and long.
Availability is thin. You can't yet walk into most stores and pick a sodium-ion power station off the shelf the way you can with LFP. Selection is limited.
The cost advantage isn't fully realized. LFP is already so cheap that Na-ion's theoretical cost edge won't show up at the register until manufacturing scale catches up.

What it means if you're buying in 2026

For most people, right now: LFP is still the correct default. It's proven, cheap, widely available, and long-lived. Don't wait for sodium-ion if you need power today.
If you live/travel somewhere genuinely cold: sodium-ion is the one to watch. Its cold tolerance is a real, structural advantage over LFP — not marketing. As units become available, cold-climate off-gridders may be the first to switch.
Watch the budget tier: if Na-ion's cost advantage materializes at scale, the cheapest power stations a few years out could be sodium-ion, not LFP.

My read: this isn't a comeback for NMC — it's a new axis of competition. LFP keeps the mainstream for a while, sodium-ion carves out cold-climate and eventually budget segments, and the two likely coexist rather than one wiping out the other.

For the data/dev crowd

I track this stuff in a structured, open dataset (specs, $/Wh, cycle life, chemistry) with a free JSON API — handy if you want to slice the market yourself or have an AI agent query it:

Dataset + data study: https://sunsee.cc/data
Free API: https://sunsee.cc/api/products.json
Sizing calculator (now with a cold-climate setting, since cold is half this story): https://sunsee.cc/calculator

If you've gotten hands-on with a sodium-ion cell or power station, I'd love to hear real-world numbers in the comments — especially cold-weather behavior and cycle life.

Data from SunSee (https://sunsee.cc), CC-BY-4.0.

What 102 Portable Power Stations Tell Us About Buying One in 2026

hardy lee — Mon, 03 Aug 2026 12:26:24 +0000

If you've ever tried to buy a portable power station, you know the problem: every brand claims to be the best, the spec sheets are a wall of numbers, and the forums are full of confident but contradictory advice. "What size do I actually need?" is the most-asked question and the least-clearly-answered.

So I did the boring thing. I built a structured database of 102 portable power stations from 24 brands — capacity, output, chemistry, cycle life, solar input, weight, price — and started running the numbers. A few findings were genuinely surprising.

1. Prices quietly collapsed

The median portable power station now sits at $0.61 per watt-hour. The cheapest in the dataset is $0.39/Wh (the GRECELL T1000). A few years ago, ~$1/Wh was normal and anything under $0.70 felt like a deal.

The practical takeaway: if you're paying much more than ~$0.70/Wh in 2026, you're mostly paying a brand premium. That premium sometimes buys you a better app, ecosystem, or support — but it's worth knowing you're paying it.

2. LiFePO4 basically won

94% of the models I track now use LiFePO4 (lithium iron phosphate) instead of the older lithium-ion (NMC) chemistry. This matters more than any marketing bullet point:

LiFePO4: ~3,000–4,000+ charge cycles
Older Li-ion (NMC): ~500 cycles

At daily use — say you cycle it every day in a van or for backup — that's roughly 8 years vs 18 months before the battery is meaningfully degraded. If a listing still uses NMC to hit a lower price, that "deal" can cost you far more over its life.

3. The fridge myth costs people hundreds of dollars

This is the single most common sizing mistake. People size a giant, expensive battery to run a fridge because they do the math like this:

Fridge nameplate (150W) × 24 hours = 3,600 Wh/day

But a fridge's compressor only runs about 40% of the time. Its real average draw is closer to 60W, so:

150W × 40% × 24h ≈ 1,440 Wh/day

In runtime terms: a 1,000Wh power station runs a full-size fridge for about 14 hours, not the ~6 hours the nameplate math implies. Sizing on the nameplate number pushes people into a battery twice as big — and twice as expensive — as they need.

The same logic applies to anything with a compressor or thermostat: fridges, coolers, some medical devices.

4. Small devices are cheaper to run than you think

A basic CPAP machine (no heated humidifier) sips about 40W, so a modest 500Wh station gives you a full night. Flip on the heated humidifier and the draw roughly doubles — now you want 750Wh+ for the same night. Knowing which features double your power draw is worth more than buying a bigger battery.

How to size your own system (the honest formula)

You don't need a huge battery; you need the right one. Here's the math the calculator on my site automates:

Daily energy: add up each device's watts × hours per day (remember the ~40% duty cycle for fridges/coolers).
Battery size: daily Wh × backup days ÷ 0.85 (LiFePO4 gives you ~85% usable after real-world losses).
Solar to recharge in a day: daily Wh ÷ peak sun hours ÷ 0.75 (the 0.75 accounts for wiring, controller, and temperature losses).
Inverter: add the running watts of everything you'd run at once, then add ~25% headroom for startup surge.

If you want it done automatically from an appliance list, I built a free calculator here: https://sunsee.cc/calculator

Methodology & the full data

Everything above comes from a spec database I maintain. To keep product comparisons honest, I score each unit 0–100 across six weighted dimensions — value ($/Wh), cycle life, power density, charge speed, solar capability, and output — with the formula published openly (no pay-to-play):

Full data study, with per-brand $/Wh and price-by-year tables: https://sunsee.cc/data
Scoring methodology: https://sunsee.cc/score

The data is free to cite (CC-BY) — just credit the source.

How it's built (for the devs)

Under the hood this is a static site: Astro + React + Tailwind on Cloudflare Pages, $0/month hosting. The interesting bet is on distribution: instead of only chasing Google rankings, everything is exposed as a free JSON API + an MCP discovery endpoint + an llms.txt, so AI assistants can query products, run the sizing calculator, and fetch runtime data directly — with attribution and affiliate links included.

The theory: a growing share of "which power station should I buy?" questions will be answered by an assistant, not a search box. So the goal is to be the data source those assistants cite, not just another blog fighting for page one.

API: https://sunsee.cc/api/products.json
Discovery: https://sunsee.cc/api/mcp.json

If you've measured referral traffic from AI assistants — or have opinions on backlinks vs classic SEO for a data/tool site — I'd love to hear it in the comments.

Data from SunSee (https://sunsee.cc), CC-BY-4.0. Prices are MSRP and may vary by retailer.