PHIUTO
HomeProductCompanyInvestors
Get in touch
Product

A pipeline that builds itself around your goal.

Cost and expected coverage before execution; strategy changes until the criteria are met — or an honest account of why they can't be.

Plan Mode

See the cost before it is spent.

Phiuto weighs acquisition strategies across public, customer-connected and commercial sources — scored on coverage, quality, freshness, cost, speed, access and licensing. It picks the best strategy for your objective, not the most complete one available: a fast directional answer should not be sold a €300 path.

StrategyCoverage CostConfidence
Single public source
72%€081%
Multiple public sources
89%€088%
Public + commercial API
96%€30095%
Illustrative example — not measured results.

If a goal needs paid access, a credential or a licence decision, Phiuto asks rather than assuming.

Why it's different

Goal-driven, not workflow-driven.

Traditional automation
Input
Step 1 → Step 2 → Step 3
Done

The pipeline is defined in advance. It stops when the steps are done — whether or not you got what you needed.

Phiuto
Goal — “Find 500 European manufacturers that meet these requirements.”
Plan → Execute → Evaluate
Achieved? Verify → deliver
Not yet? Re-plan → execute again ↺

The pipeline is constructed around the objective. It stops when the objective is satisfied — or is shown to be unattainable, and says so.

Traditional automation executes a workflow. Phiuto pursues an objective.

The system

A completed goal is not one API call.

Between your objective and a verified answer sits a coordinated system of specialised components — each one a real engineering problem. The agents are not the product; the system that drives them toward the objective, and can prove how it got there, is.

live trace — simulated
Orchestratorowns the objective, the budget and the stopping rules — the only component that can decide to stop
Requirements
natural language → measurable criteria; ambiguity is surfaced, never guessed
Source intelligence
what each source holds, how it's accessed, what it costs, when it breaks
Strategy
scores acquisition paths on coverage, quality, freshness, cost and licensing
Acquisition & extraction
APIs, websites, registries, documents → structured records
Entity resolution
one company = one record, with stable identifiers across sources and languages
Verification
cross-checks against genuinely independent sources; circular corroboration is detected
Goal evaluatorsatisfied → deliver · not satisfied → re-plan · unattainable within constraints → says so
Memoryevery run writes back — strategies, failures, costs, source behaviour, resolved entities. Run two starts smarter than run one.where it compounds
Non-negotiables — enforced in code
01Provenance is part of the record, not a log.
02Confidence is computed from corroboration — never asserted by a model.
03Licence terms are checked in code, at acquisition and at delivery.
04Budgets are hard ceilings — nothing spends past your approval.
05Every run is replayable — auditable end to end.
06“Could not obtain X, because Y” is an output type, not an exception.

The models are rented. The source intelligence, the entity graph and the evaluation data are ours — and they compound with every run.

The source graph

The world's data, as a territory our agents navigate.

Every source we've ever touched — registries, APIs, filings, sites — is a node in a graph, connected by what links to what. Agents traverse it in parallel: reaching a node probes it, a hit turns it red and opens its neighbours, a dead end is recorded so no agent wastes time there again. The graph is live below.

Source intelligence graph — live traversalhitprobedunknown
Agents
6
Sources probed
1
Hits
0
Dead ends mapped
0

Every traversal leaves the graph smarter: which sources answer, what they cost, where they fail. That map is the asset no competitor can copy by calling the same APIs.

In depth

If you want the detail.

The core is above. The rest is here — positioning, trust, pricing — for whoever is looking for it.

What Phiuto is not+

Not a database. We don't aim to own data — we aim to know how to reach it, and to be the layer that reaches it well.

Not a scraper. Scraping is one acquisition method among many, and usually not the first we reach for.

Not a workflow builder. No canvas of nodes to wire up. If you have to design the pipeline, we've failed at the core premise.

Not a general-purpose agent. Phiuto does one thing: acquire external data against an objective. It does not write your emails.

Not a model company. The model is a component. The intelligence is in source knowledge, strategy selection and evaluation.

Not a replacement for the analyst. Phiuto assembles and verifies the information; the human decides.

Trust, provenance and licensing+

For every material result, Phiuto can expose its source, retrieval time, transformation history, verification, confidence — and its limitations: what remains uncertain or unobtainable.

Licence terms are enforced in code: retention, redistribution and rate limits are attributes of each source, checked at acquisition and at delivery. Budgets are hard ceilings — no goal spends past its authorised limit without a checkpoint.

The goal isn't to pretend data is perfect. The goal is to make data — and its uncertainty — transparent. In procurement and financial contexts, an unsourced number is unusable regardless of whether it's correct.

How pricing works+

Priced per completed goal where possible — it matches the product's own unit of execution, and it's the only pricing you can evaluate before spending. Plan Mode shows estimated cost and expected coverage before execution.

Four layers, sequenced: SaaS for teams through a focused workflow · usage per completed goal · a developer API for applications and AI agents · enterprise infrastructure for high-volume acquisition. Land with workflows; expand into infrastructure.

Questions we get+

Isn't this just scraping?

Scraping is one method and usually not the first. The product is strategy selection and completion measurement.

How do you know the data is right?

Confidence is computed from independent corroboration, not asserted, with provenance on every material field. We'd rather return 82% with a gap report than 100% that's quietly wrong.

Won't the AI labs just do this?

Models improve reasoning, not access. Licences, credentials, source behaviour, entity resolution and verification are accumulated operational assets.

What happens when a goal can't be completed?

Phiuto says so: what it could not obtain, why, and what it would cost to close the gap. A system that can only succeed will fabricate.

ContinueCompany