Panel Database API

A REST API over the panel database used by the FirstPoint Solar Panel String Sizer: roughly 1,300 models with STC and NOCT ratings, temperature coefficients, bifaciality, and a direct download link to the manufacturer datasheet wherever one is on file. Partners can also add panels by uploading a datasheet: our AI extracts every variant, you review the result, and nothing is saved until you say so.

All responses are JSON (UTF-8). Successful responses put the payload under data; list responses add a meta object for paging. The API is versioned in the path — breaking changes will ship as /api/v2, and v1 will keep working. New fields may be added to v1 responses at any time, so ignore keys you do not recognise.

Authentication

Every panel request needs an API key. Contact FirstPoint Energy to request one. Send it in either of these headers:

Authorization: Bearer fpk_…
# or
X-API-Key: fpk_…

Requests without a valid key get 401. Keep the key server-side: the API allows cross-origin browser requests (CORS *), but a key embedded in public JavaScript is visible to anyone and will be revoked if abused.

Endpoints

Returns a page of panels. All parameters are optional and combine with AND.

The response carries the page under data and paging info under meta. Keep requesting with offset=meta.nextOffset until it is null.

{
  "data": [ { "id": 225, "name": "Canadian Solar CS6.1-54TM-450H", … } ],
  "meta": { "total": 8, "limit": 100, "offset": 0, "nextOffset": null }
}

Returns a single panel by its numeric id, wrapped in data. Unknown ids return 404.

The machine-readable OpenAPI 3.1 description of this API — import it into Postman, Insomnia, or a client generator. GET /api/v1 (no key) returns a short index with links to the endpoints and this page.

The panel object

{
  "id": 16,
  "name": "Alexus Solar ALEX-400-B-54-S",
  "verified": true,
  "bifacial": false,
  "stc":  { "watts": 400, "voc": 37.21, "vmp": 31.18, "imp": 12.83, "isc": 13.67 },
  "noct": { "watts": 302.3, "imp": 10.58, "isc": 11.04 },
  "bifacialWatts": null,
  "ptcWatts": 370,
  "tempCoeff": { "voc": -0.26, "pmp": -0.34, "isc": 0.05 },
  "datasheetUrl": "https://1jlpnihnya4zwkzy.public.blob.vercel-storage.com/datasheets/16.pdf",
  "createdAt": "2026-08-18T22:20:20.403Z"
}

Datasheets

datasheetUrl is a plain HTTPS link to the PDF on FirstPoint's file storage. It needs no API key, so you can hand it straight to end users or fetch it server-side. Most files are the manufacturer's original datasheet; a few are archived copies where the manufacturer's link has since gone dead. Filter with hasDatasheet=true when you only want panels that have one.

Adding panels

Adding a panel is a three-step flow so that nothing lands in the shared database without a human (or your own validation code) looking at it first. Every saved panel is linked to the archived datasheet and starts as verified: false until a FirstPoint reviewer checks it.

1 · Upload the datasheet. Send the PDF in whichever way is convenient:

# 1a. ask for somewhere to put it (filename is optional)
curl -X POST -H "Authorization: Bearer $FP_API_KEY" -H "Content-Type: application/json" \
  -d '{"filename":"JAM60-500LB.pdf"}' \
  https://string-sizer.vercel.app/api/v1/extractions/uploads

# → {"data":{"uploadUrl":"https://…","datasheetUrl":"https://…","expiresAt":"…","maxBytes":26214400}}

# 1b. PUT the PDF to uploadUrl — no auth header, the URL is the credential
curl -X PUT -H "Content-Type: application/pdf" --data-binary @JAM60-500LB.pdf "$UPLOAD_URL"

# 1c. then extract from it
curl -X POST -H "Authorization: Bearer $FP_API_KEY" -H "Content-Type: application/json" \
  -d "{\"datasheetUrl\":\"$DATASHEET_URL\"}" \
  https://string-sizer.vercel.app/api/v1/extractions

The PDF (max 25 MB) goes through our AI extractor, which reads every power class in the datasheet's electrical tables. Expect the call to take 20–60 seconds. The response is an extraction: an id, the archived datasheet URL, and one draft per variant. Nothing has been saved yet.

{
  "data": {
    "id": "ext_Qm9vZ2llV29vZ2ll",
    "status": "pending",
    "filename": "CS6.1-54TM datasheet.pdf",
    "datasheetUrl": "https://….public.blob.vercel-storage.com/datasheets/uploads/…pdf",
    "panels": [
      { "index": 0, "name": "Canadian Solar CS6.1-54TM-450H", "brand": "Canadian Solar",
        "website": "csisolar.com", "bifacial": false,
        "stc":  { "watts": 450, "voc": 41.5, "vmp": 34.9, "imp": 12.9, "isc": 13.6 },
        "noct": { "watts": 338, "imp": 9.75, "isc": 10.9 },
        "bifacialWatts": null, "ptcWatts": null,
        "tempCoeff": { "voc": -0.26, "pmp": -0.34, "isc": 0.05 } },
      { "index": 1, "name": "Canadian Solar CS6.1-54TM-455H", … }
    ],
    "savedPanelIds": [],
    "createdAt": "2026-09-04T22:41:07.512Z",
    "expiresAt": "2026-09-05T22:41:07.512Z",
    "savedAt": null
  }
}

2 · Review. Check the drafts against the datasheet — the AI is accurate on clean tables but can misread scanned or unusual layouts. You can fetch the extraction again at any time within 24 hours (only with the key that created it), which lets a separate reviewer or job do the checking. Anything you want to correct goes into the save command; the extraction itself is immutable.

3 · Save. The explicit command that writes to the database. Send no body to save every draft exactly as extracted, or choose drafts by index and override any reviewable field: name, bifacial, stc, noct (an object or null), bifacialWatts, ptcWatts, tempCoeff. Objects merge field by field, so { "stc": { "voc": 41.6 } } changes only Voc.

{
  "panels": [
    { "index": 0 },
    { "index": 1, "name": "Canadian Solar CS6.1-54TM-455H", "stc": { "voc": 41.9 }, "ptcWatts": 412 }
  ]
}

The response lists the panels that were created (with their new ids, ready to use in GET /panels/{id}) and any that were skipped because a panel with the same name already exists — existing entries are never overwritten. Each extraction can be saved once; a second save returns 409, and an extraction older than 24 hours returns 410.

{
  "data": {
    "extraction": { "id": "ext_Qm9vZ2llV29vZ2ll", "status": "saved", "savedPanelIds": [1312], … },
    "saved":   [ { "id": 1312, "name": "Canadian Solar CS6.1-54TM-455H", "verified": false, … } ],
    "skipped": [ { "index": 0, "name": "Canadian Solar CS6.1-54TM-450H", "reason": "duplicate",
                   "message": "A panel with this name already exists." } ]
  }
}

The whole flow with curl

# 1. upload (datasheets over ~3 MB: see the two-step route above)
curl -X POST -H "Authorization: Bearer $FP_API_KEY" \
  -H "Content-Type: application/pdf" -H "X-Filename: CS6.1-54TM.pdf" \
  --data-binary @CS6.1-54TM.pdf \
  https://string-sizer.vercel.app/api/v1/extractions

# 2. review the JSON above, then

# 3. save everything as extracted …
curl -X POST -H "Authorization: Bearer $FP_API_KEY" \
  https://string-sizer.vercel.app/api/v1/extractions/ext_Qm9vZ2llV29vZ2ll/save

# … or only the variants you checked, with corrections
curl -X POST -H "Authorization: Bearer $FP_API_KEY" -H "Content-Type: application/json" \
  -d '{"panels":[{"index":1,"stc":{"voc":41.9}}]}' \
  https://string-sizer.vercel.app/api/v1/extractions/ext_Qm9vZ2llV29vZ2ll/save

MCP server

The same database, extraction flow and the String Sizer's own sizing engine are also available as a Model Context Protocol server, so AI assistants and agents can use them directly. It speaks Streamable HTTP, is stateless, and authenticates with the same API key sent as a bearer token.

Claude Code

claude mcp add --transport http firstpoint-panels https://string-sizer.vercel.app/api/mcp \
  --header "Authorization: Bearer $FP_API_KEY"

Cursor, Windsurf and other JSON-configured clients

{
  "mcpServers": {
    "firstpoint-panels": {
      "url": "https://string-sizer.vercel.app/api/mcp",
      "headers": { "Authorization": "Bearer fpk_…" }
    }
  }
}

Claude Messages API (MCP connector)

{
  "model": "claude-opus-5",
  "max_tokens": 16000,
  "mcp_servers": [{
    "type": "url",
    "url": "https://string-sizer.vercel.app/api/mcp",
    "name": "firstpoint-panels",
    "authorization_token": "fpk_…"
  }],
  "tools": [{ "type": "mcp_toolset", "mcp_server_name": "firstpoint-panels" }],
  "messages": [{ "role": "user", "content": "How many Canadian Solar 450 W panels per string on a FlexBOSS21 in Calgary?" }]
}

Send the header anthropic-beta: mcp-client-2025-11-20 with that request. Clients that only support OAuth-based remote servers (for example the custom-connector directory in the Claude apps) are not supported yet; bearer-token clients are.

Extraction through MCP takes the same 20–60 seconds as the REST call, and the same terms apply: the model is reading transcribed data, not a verified design.

Errors

Errors use standard HTTP status codes and a JSON body with a stable code and a human-readable message.

{ "error": { "code": "bad_request", "message": "limit must be at most 500" } }

Examples

Search with curl

curl -H "Authorization: Bearer $FP_API_KEY" \
  "https://string-sizer.vercel.app/api/v1/panels?q=rec%20pure&minWatts=400&hasDatasheet=true"

Fetch one panel

curl -H "X-API-Key: $FP_API_KEY" https://string-sizer.vercel.app/api/v1/panels/16

Download every panel (JavaScript)

const BASE = 'https://string-sizer.vercel.app/api/v1';
const headers = { authorization: `Bearer ${process.env.FP_API_KEY}` };

async function allPanels() {
  const panels = [];
  for (let offset = 0; offset !== null;) {
    const res = await fetch(`${BASE}/panels?limit=500&offset=${offset}`, { headers });
    if (!res.ok) throw new Error(`${res.status}: ${(await res.json()).error.message}`);
    const { data, meta } = await res.json();
    panels.push(...data);
    offset = meta.nextOffset;
  }
  return panels;
}

Incremental sync

Remember the newest createdAt you have seen and ask only for what came after it:

GET /panels?createdAfter=2026-08-18T22:20:24.902Z&sort=createdAt&order=asc

Terms of use

Use of the API is subject to the FirstPoint Energy Terms of Use. In short: the data is provided “as is” with no warranty, it is not engineering advice, you must verify every value against the manufacturer’s current datasheet and applicable codes, and FirstPoint Energy accepts no liability for any loss arising from its use. By requesting a key you accept those terms.

Practical notes

• Data quality. Entries with verified: false were extracted automatically from an uploaded PDF and have not been reviewed — check them against the linked datasheet before relying on them for design work. Verified entries can still be wrong; verify those too.
• Caching. Panel responses are marked Cache-Control: private, max-age=60. The full dataset changes rarely; pulling it once a day is plenty.
• Fair use. No hard rate limit is enforced today. Please keep sustained traffic under a few requests per second and use paging rather than repeated full downloads; abusive keys are revoked.
• Uploads. Only upload datasheets you are entitled to share; archived copies are served publicly as datasheetUrl. Panels you save are visible to every API consumer and to String Sizer users, flagged unverified until reviewed.
• Keys. Keys are personal to the partner they were issued to, may not be shared or embedded in public code, and may be revoked at any time.
• Attribution. When you show this data to your users, please credit FirstPoint Energy and pass the disclaimer on to them.