Decide where every click goes — in under a millisecond of logic.

Vector TDS pairs a rule-based delivery engine with an OpenRTB/XML buying engine, a masking reverse proxy and an AI optimizer that classifies every traffic source and tunes bids in the background.

01 / delivery

Edge routing in one hop

Every hit is resolved from a warm in-isolate config cache and answered with a 302/307 or a masked proxy stream — no cold database round trip on the hot path.

02 / logic

Visual rule builder

Compose branches on geo, device, OS, browser, webview, bot score, time and query params. Rules compile to a JSON AST evaluated allocation-light per request.

03 / proxy

URL masking with an allow-list

The reverse-proxy engine streams the offer under your own domain, rewrites asset paths and refuses any destination host that is not explicitly allow-listed.

04 / tracking

Async logging & postbacks

Clicks are batched and flushed off the response path; conversions arrive on a secret-signed postback endpoint and roll straight into the dashboards.

Buying engine

Buy the traffic in the same system that routes it.

Vector answers OpenRTB bid requests and serves XML feeds from edge routes. Every provider gets its own signed endpoint, its own targeting and its own spend ledger.

Per-provider bid tokens — revoke one source without touching the rest.
Coarse QPS throttle in front of every endpoint keeps the bid loop predictable.
Bid, win and spend logging rolls straight into the dashboards and telemetry.
Feed templates for ZeroPark, ExoClick, Adsterra, PopAds and a generic format.

Integration surface

OpenRTB 2.5

/api/public/rtb/bid/{token}

Win notice

/api/public/rtb/win/{id}

XML feed

/api/public/xml/feed/{token}

TDS + RTB

A single loop from bid request to conversion.

bid request -> targeting + AI intent check -> bid response -> win notice
     -> click hits /r/{token} -> rule AST -> redirect or masked proxy
     -> click log -> postback conversion -> telemetry

One rule AST, two moments

The same compiled branch tree decides whether to bid on an impression and where the resulting click lands. Targeting never drifts between buying and delivery.

Closed measurement loop

Spend from the win notice and payout from the postback meet on the same source key, so ROI and EPC are known per publisher without exporting anything.

No hop left unexplained

Bid, win, click and conversion share one identifier end to end — a source that wastes budget is visible in the same pass that priced it.

AI optimization

An optimizer that never sits in the bid path.

The model runs in the background and writes into a warm in-isolate cache. If it is slow, rate-limited or down, the engine keeps bidding on the last known good rules — traffic never waits on inference.

Semantic source classifier

Publisher domains, app bundles and placement keys are sorted into iGaming/Casino, Crypto, Adult, General Entertainment or Junk/Farming — so the bid loop can target or drop a source the moment it appears.

Autonomous feedback loop

Telemetry rolls up per source every few minutes: impressions, clicks, conversions, spend, payout, EPC and ROI. The model returns blacklists and bid multipliers that apply instantly and stay reversible.

Rules you approve

New IF/ELSE routing logic never applies itself. It lands in the workspace as a proposal with its rationale and the exact branch JSON, waiting for a human OK.

Natural-language command center

“Filter ExoClick for iGaming intent only, and raise CPM 15% on placements above 30% ROI” becomes concrete blocks, multipliers and rule proposals.

Try the intent classifier

Enter a publisher domain or app bundle. This runs the live classifier and stores nothing.