Makita Deal Finder — Austria

Parallelism Strategy

Always launch multiple background subagents with different AI models to maximize coverage. Different models find different products because they use different search strategies and interpret web results differently.

Use the task tool with mode: "background" to launch agents in parallel:

Agent A: model "claude-opus-4.6" → search for Makita LXT + XGT solo on geizhals.at
Agent B: model "gpt-5.4"         → same task, different model for diverse results

Start Phase 2 (Geizhals API calls) immediately using the known product ID map below — don't wait for discovery agents to finish. When they complete, pull history for any NEW products they found.

Search Strategy

Each discovery agent should:

Agent prompt template:

Search for Makita [18V LXT / 40V XGT] solo tools available in Austria on geizhals.at. For each tool found, collect: Model number, description, Geizhals product ID (from URL, e.g., a3640795.html → ID 3640795), current lowest price (EUR), and category. Only SOLO tools — no battery packs. Try web searches like site:geizhals.at makita [LXT/XGT] solo, search specific categories on geizhals.at, and search for known popular models.

Known Product ID Map (verified April 2026)

18V LXT tools:

DTD173Z=3368691  DHP487Z=2586050  DJV185Z=2987632  DJV180Z=1193744  DTM52Z=2511437
DTW300Z=2315782  DHP484Z=1552629  DTD152Z=1481309  DBO180Z=1171542  DTW181Z=2012354
DKP180Z=1238531  DGA504Z=1335263  DHR243Z=1315665  DSS611Z=1222079  DHR171Z=1722508
DDF484Z=1693050  DJR186Z=1647641  DGA513Z=1834969  DRT50Z=1907804   DHR182Z=2137721
DHS680Z=1330684  DHP486Z=2560455  DCL182Z=1217596  DCL180Z=1460400  DST221Z=1786110
DKP181Z=2285099  DUB187Z=2897266

40V XGT tools:

UC030GZ03=3640795  HR004GZ01=3758528  UN001GZ=3504669    TW004GZ01=3036789
TM001GZ=3319438    GD001GZ=3646971    DF001GZ=2349569    VC004GLZ01=3218741
RP001GZ=3333156    HR010GZ=3105395    UC016GZ=2736532    TD002GZ01=3033865
SP001GZ03=2665284  CE003GZ02=3332803  AS001GZ=2717000    GA005GZ=2349758
GA023GZ=2349789    GA038GZ=2557696    HR002GZ=2349476    HP001GZ01=2349222
DF002GZ=2564837    HR005GZ01=2558085  JV002GZ=3149723    UC011GZ=2736471
KP001GZ=2745034    HR007GZ=2720981    JR002GZ=2739628    GA047GZ=2820665
HP002GZ=2557591    HS012GZ=2930595    TD001GZ=2349641    UC002GZ=2736559
MP001GZ=2690688    CF001GZ=2460555    CL001GZ02=2631553  HS004GZ=2360545

Geizhals ID Verification

CRITICAL: Geizhals IDs from search agents are often wrong (mapping to bicycle tires, socket wrenches, etc.). Always verify IDs before pulling price history by either:

• Checking the product page URL returns a Makita tool (not some other product)
• Confirming the price from the API is in a plausible range for a power tool (€50-€1000)
• Cross-referencing with web search: site:geizhals.at makita <MODEL> solo

Red flags for wrong IDs: Price under €20, price over €2000, product name mentioning tires/sockets/unrelated items, ATL of €0.01 or similar anomalies.

Phase 2: Pull Price History from Geizhals API

API Endpoint (discovered via reverse engineering)

POST https://geizhals.at/api/gh0/price_history

Request body:

{"id": [PRODUCT_ID], "params": {"days": 9999, "loc": "at"}}

• id: Array of Geizhals product IDs (integers). Supports multiple products per call.
• days: Must be one of: 7, 31, 91, 183, 365, 9999 (full history). Use 9999 for ATL analysis.
• loc: "at" for Austria, "de" for Germany.

Required headers:

Content-Type: application/json
Origin: https://geizhals.at
Referer: https://geizhals.at/

Response format:

{"response": [[timestamp_ms, price, offer_count], ...], "meta": {...}}

• timestamp_ms: Unix epoch in milliseconds
• price: Lowest available price in EUR on that day (null = no offers)
• offer_count: Number of shops offering the product

PowerShell Script Pattern

$ids = @{ 'ModelName'=GeizhalsID; ... }
$headers = @{
    'Content-Type'='application/json'
    'Origin'='https://geizhals.at'
    'Referer'='https://geizhals.at/'
}
$results = @()
foreach ($kv in $ids.GetEnumerator()) {
    $body = "{`"id`":[$($kv.Value)],`"params`":{`"days`":9999,`"loc`":`"at`"}}"
    try {
        $resp = Invoke-RestMethod -Uri 'https://geizhals.at/api/gh0/price_history' -Method Post -Body $body -Headers $headers
        $data = $resp.response | Where-Object { $_[1] -ne $null }
        if ($data.Count -gt 0) {
            $sorted = $data | Sort-Object { $_[1] }
            $atl = $sorted[0][1]
            $atlDate = [DateTimeOffset]::FromUnixTimeMilliseconds($sorted[0][0]).DateTime.ToString('yyyy-MM-dd')
            $current = $data[-1][1]
            $pct = [math]::Round(($current - $atl) / $atl * 100, 1)
            $results += [PSCustomObject]@{
                Model=$kv.Key; Current=$current; ATL=$atl;
                ATLDate=$atlDate; VsATL="+${pct}%"; GhID=$kv.Value
            }
        }
    } catch { Write-Host "FAIL: $($kv.Key)" }
    Start-Sleep -Milliseconds 300  # Rate limit courtesy
}
$results | Sort-Object {
    [double]($_.VsATL -replace '[+%]','')
} | Format-Table -AutoSize

Data Quality Notes

• ATLs from 2019-2022 were often set during summer sales (June-July) or Black Friday
• Some ATLs are pricing errors (e.g., €0.01, €14.81 for a blower) — flag anything under €20 as anomalous
• Null price values in the response indicate days with no offers — filter these out
• The meta field in the response is unexplored and may contain useful shop-level data

Phase 3: Compare with Additional Sources

Launch these as parallel background subagents — one per source — so they run concurrently with each other and with any remaining Phase 2 work:

Agent C: "willhaben search"  → search willhaben.at for NEW Makita solo tools
Agent D: "eBay search"       → search eBay.de for NEW Makita solo tools shipping to AT

Both agents should search for the top deal candidates from Phase 2 (especially tools near ATL) plus broad searches like "Makita 18V solo" and "Makita XGT solo".

willhaben.at (Austrian marketplace)

Search for NEW Makita tools. Filter out used items.