On-chain evidence your team can verify.

Self-hosted DeFi analytics for funds, protocols and risk teams. Explore any token across chains, catch unusual activity with the evidence attached, publish source-backed reports, and ask an AI analyst that only states numbers it can prove.

Request a Private Pilot See what's verified

Runs on your own servers. Your organisation's data and AI prompts stay there unless you choose a cloud AI provider.

Anomaly center in a desktop browser: high-severity findings for VELO, XDP and other subjects with network, window, observed-to-baseline ratio and evidence strength

Token explorer on a tablet: VELO on OP Mainnet verified on-chain, with twelve pools each marked supported

Conversational analyst on a phone: a question about ARB on Arbitrum answered with cited facts

Real screens from the self-hosted web application on verified on-chain data (snapshot 2026-10-17), shown in a browser on each device. DeFi Intelligence is a responsive web app, not a native mobile app.

Built on receipt-verified data

Swaps are read from public JSON-RPC or your own endpoints, proven complete for every block range and cross-checked before anything is published. A protocol family is supported only after its decoded swaps matched transaction receipts exactly on several networks. Uniswap v4 is verified differently: its pools share one contract and can carry hooks, so each pool is verified on its own, and only individually verified pools are ingested.

Networks: BaseArbitrum OneOP MainnetEthereum
Verified protocol families: Uniswap v3-styleUniswap v2-styleSolidly (Velodrome/Aerodrome v2)
Individually verified pools: Uniswap v4: six Ethereum 1INCH pools
Not counted: PancakeSwap v3 (decoder only, not verified for ingestion)Other Uniswap v4 pools and networksv4 pools with native ETH or swap-changing hooksOther venues

For teams whose numbers have to hold up

Every figure carries its coverage, its valuation method and the SQL behind it, so analysis can be reproduced, reviewed and shared.

Protocols, DAOs and token teams

See how your token trades across chains and pools — volume, liquidity and valuation coverage — and get flagged when activity departs from its baseline.

Funds and research desks

Source-backed token and protocol research with stated coverage, so a conclusion can be traced back to the swaps that produced it.

Risk and market-integrity analytics

Anomaly findings for investigation triage: statistical observations with baselines, limitations and alternative explanations, never accusations.

Analytics consultancies

Client-ready reports from verified data on private infrastructure, with an AI analyst that answers in several languages without inventing numbers.

How it works

Three stages, each with its own checks. Nothing reaches a report or an AI answer without passing the stage before it.

Verified data
Ingestion runs prove block coverage, confirm empty answers with a second provider and publish to ClickHouse only after quality gates pass. USD comes from stablecoin legs or verified hourly prices; unvalued swaps are never counted as $0.
Analytics and findings
Covered-hour metrics, window comparisons and an anomaly engine with frozen baselines. Reports state exactly which pools and hours they cover.
Evidence-grounded AI
The AI phrases answers from the verified facts of each question. Every number must match a fact the sentence cites; anything unproven is replaced by a clearly labelled deterministic answer.

One workflow from token to report

Start with any token contract on a supported network and end with findings and a report you can defend.

Universal token explorer

Paste a token contract. DeFi Intelligence checks its identity on-chain and against an independent registry, discovers its pools and tells you which ones it can analyse.

A support level and reason for every pool
Collects only the hours that are missing, through the same gated pipeline
Per-pool coverage, quality checks and an evidence-backed report

Token explorer listing twelve VELO pools on OP Mainnet with family, pair, liquidity and a supported label for each

Anomaly intelligence

Volume, activity and liquidity shifts against a baseline frozen at onset, plus concentration, imbalance, divergence and repetitive patterns.

Facts, method and parameters for every finding
Limitations and alternative explanations, stated up front
Statistical observations — never claims of intent or manipulation

A VELO volume spike finding: chart with frozen baseline and observation window, facts, method, limitations, alternative explanations and provenance

Research and reporting

Reports in Markdown, HTML and JSON with price-source coverage, pool coverage, protocol breakdown, multi-hop reconciliation and the SQL evidence.

Window comparisons marked comparable only when every pool is covered in both
Workspaces with saved, frozen reports per organisation
Pool-event volume and transaction-net volume reported separately

Historical comparison of two ARB windows on Arbitrum One, both fully covered, with swap events, token volume, valued USD volume and changes

An AI analyst that cites its sources

Ask in English, Turkish or German and keep the conversation going: follow-up questions keep the token and topic. Each answer is labelled either as a validated AI answer or as a deterministic answer built from the same verified facts.

Inference option	Status	Notes
Local models via Ollama	Tested	Runs on your hardware; a memory guard prevents overloading the host
Anthropic Claude API	Integrated	Optional and off by default; no paid call until an administrator authorises it within limits. Live calls not yet verified
OpenAI API	Integrated	Same controls as Claude. Live calls not yet verified
Self-hosted OpenAI-compatible servers	Integration option	vLLM, llama.cpp, LM Studio or an organisation gateway through the same adapter; GPU serving not yet tested

Local model benchmark

The same frozen fact set (10 questions in English, Turkish and German; 8 answered by the model, 2 declined by deterministic pre-checks) and the same validators, on a 16 GB machine.

Model	Accepted by validation	Grounding failures	Avg latency	Memory
gemma3:4b	8 / 8	0	8.3 s	2.9 GB
qwen3.5:9b	6 / 8	2	22.4 s	5.7 GB

In release testing on 2026-10-17, gemma3:4b answers to an English question and English and Turkish follow-ups were all accepted. In a later run a Turkish answer that cited the wrong facts was rejected and replaced by the deterministic answer — exactly what the validation is for. One machine, not a performance promise.

The analyst's answer about ARB on Arbitrum, with numbered fact citations after each sentence

Verified research examples

Reviewed snapshots from historical, ingested windows. Findings are statistical observations: they never establish intent, manipulation or wash trading, and addresses are not identified entities.

Historical research snapshot

1INCH token on Ethereum

16 Oct 2026, 19:00 UTC to 17 Oct 2026, 19:00 UTC · 11 ingested pools, including six individually verified Uniswap v4 pools · reviewed snapshot, not live

Every Swap event emitted by the eleven selected pools in this window. These are selected-pool event measurements: not the total 1INCH market and not transaction-net trading volume.

289: pool Swap events
250: distinct transactions
24: fully covered hours, across all 11 selected pools
288 / 289: swaps valued in USD (99.7%)
$108,733.34: USD-valued pool-event volume
1: unvalued wstETH swap: no verified hourly price, so it is left unvalued, never counted as $0
28: transactions that passed through more than one ingested pool
0: anomaly findings for this window under the existing detection method

The application's token report for 1INCH on Ethereum: window 16 Oct 19:00 to 17 Oct 19:00 UTC, 24 verified covered hours, USD scope limited to the 11 ingested pools, 289 pool swaps, 1,180,334 1INCH, $108.7K valued with 288 of 289 swaps valued and 1 unvalued, 11 verified pools, and an hourly chart of 1INCH bought and sold

How to read these numbers

Selected pools, not the market. Only the eleven ingested pools are measured. Other DEX pools, other venues and centralised exchanges are not included, so this is not total 1INCH trading volume.
Pool events, not net trades. A trade routed through two ingested pools is counted in both. 28 transactions did this, which is why pool-event volume (1,180,334 1INCH) is higher than transaction-net volume (855,100 1INCH).
Valuation. 3 swaps were valued from a stablecoin leg and 285 from verified DefiLlama hourly prices. The remaining wstETH swap had no verified price for its hour and stays unvalued.
No findings. The anomaly detectors returned nothing for this single 24-hour window. That is a statistical result for a short window, not a statement about intent or market quality.

The eleven ingested pools and their Swap events in the window
Pair	Protocol and identifier	Fee tier	Swap events
1INCH / CRV	Uniswap v4 `PoolId 0x6bf6…d1fe`	1%	79
1INCH / AAVE	Uniswap v4 `PoolId 0x32dd…3a7a`	1%	50
1INCH / LINK	Uniswap v3 `Pool 0x2802…e4c2`	0.3%	47
1INCH / wstETH	Uniswap v4 `PoolId 0x1d6e…ea6f`	1%	39
1INCH / LINK	Uniswap v4 `PoolId 0x17bc…2958`	1%	33
1INCH / WETH	Uniswap v3 `Pool 0xe931…cb97`	1%	21
1INCH / WETH	Uniswap v2 `Pool 0x26aa…86f2`	0.3%	13
1INCH / WETH	Uniswap v3 `Pool 0xd35e…4b44`	0.3%	3
1INCH / USDC	Uniswap v4 `PoolId 0x1ff1…e81d`	1%	2
1INCH / WBTC	Uniswap v4 `PoolId 0x6255…d826`	1%	1
1INCH / USDC	Uniswap v3 `Pool 0x9feb…82ae`	1%	1
11 pools	6 Uniswap v4 (204 events) · 4 Uniswap v3 · 1 Uniswap v2		289

How the six Uniswap v4 pools were verified

PoolId matches PoolKey. Each pool's Initialize event, found on one RPC provider and confirmed by three others, gives its PoolKey, and the hash of that key equals the PoolId. All six are hookless 1% pools without native ETH.
Two providers agree. The 204 v4 Swap events in this window were fetched independently from two RPC providers. Both returned the identical set, and the published rows match it event for event.
Sampled receipt reconciliation. 149 sampled transaction receipts, from this window and earlier blocks, reconciled exactly with the PoolManager's token transfers (24–25 per pool), with 0 mismatches. Not every one of the 204 swaps was individually receipt-reconciled.

What is not covered

Other v4 pools are not automatically supported. Other PoolIds, Uniswap v4 on other networks, pools with hooks that can change swap amounts and pools with native ETH each need the same verification first.
PancakeSwap v3 remains decoder-only. Its swaps can be decoded, but it is not verified for ingestion.
Not the Base case. This case covers the 1INCH ERC-20 token in selected Ethereum pools. It is separate from the audited aggregator case on Base below, which analyses 1inch aggregator volume on another network.

Aggregator volume anomaly on Base

1 Sep to 15 Oct 2026, audited

3,293,676 decoded events. Window volume $458.56M, with 90.2% from one automated account; peak 10.3× the pre-onset median; volume excluding the identified account −15.5% against baseline (corrected by the audit from −20.5%).

ARB on Arbitrum One

14 Oct 12:00 to 17 Oct 03:00 UTC, 3 pools

63,089 pool swap events and $17.06M valued volume: 36.8% through stablecoin legs, 63.2% through verified hourly prices, 0 unvalued. Pass-through routing 0.1%.

VELO on OP Mainnet

16 Oct 12:00 to 17 Oct 00:00 UTC

12 pools discovered and receipt-verified; 6 ingested. Solidly pools carried 62,302 swap events but only 463,647 VELO; v3 pools carried 3,266 events and 12.2M VELO. The report states that market coverage is partial.

What's verified, and what isn't yet

We separate what has been tested on real data from what is integrated but not yet independently verified.

Capability	Status
Ingestion, token explorer, anomaly engine, reports and comparisons on Base, Arbitrum One, OP Mainnet and Ethereum	Tested on real data
Uniswap v4: six individually verified Ethereum 1INCH pools (PoolId/PoolKey match, two-provider agreement, sampled receipt reconciliation)	Tested on real data
Other Uniswap v4 pools, other networks, hooks that change swap amounts, native ETH	Not supported
PancakeSwap v3 ingestion	Decoder only
Local models via Ollama, including a live multilingual multi-turn conversation	Tested
Self-hosted Docker Compose installation: TLS proxy, first administrator, your own RPC endpoints, upgrade and rollback	Tested on Docker Desktop
Organisation roles and isolation of organisation data (workspaces, reports, members, settings, audit)	Tested
Local backups with verified test restore	Tested
Live Anthropic Claude API and OpenAI API calls	Not yet verified
vLLM on GPU servers	Not yet verified
Kubernetes deployment on a single-node test cluster	Validated locally
Encrypted S3-compatible backups with isolated restore verification, against local S3-compatible storage	Tested locally
Off-site recovery from an independent external S3 destination	Not yet verified

Deploy it on your own infrastructure

Only the TLS proxy is exposed. ClickHouse, pipelines and inference stay on internal networks, and secrets are encrypted at rest.

Profile	Status	Target	Notes
Business server	Tested on Docker Desktop	Linux with Docker Compose, 32 GB RAM recommended, 50 GB+ persistent storage, optional NVIDIA GPU	Read-only preflight installer, first-administrator setup, upgrade and rollback by release tag
Local evaluation	Tested	A 16 GB machine	Small local models, memory guard, automatic ingestion off by default
Enterprise	Validated locally	Kubernetes, separate GPU inference servers	Kubernetes manifests with optional Airflow. Upgrade, rollback, persistence, network policies and backups validated on a local single-node cluster; production clusters, managed cloud, external TLS ingress and GPU inference nodes not yet validated

Backups are encrypted before they leave the server. Point them at your own S3-compatible bucket for off-site copies; every backup can be restore-verified in isolation, without modifying production data. Tested end-to-end against local S3-compatible storage; recovery from an independent remote destination has not yet been verified.

Starting points, not performance promises: capacity depends on the measured hardware.

Run a private pilot on your own server

A pilot runs in evaluation mode for one organisation. Bring a token or protocol you want to analyse on a supported network; collection, findings, reports and the AI analyst run against verified on-chain data.

Request a Private Pilot

Opens a short email template: organisation, contact, use case, preferred deployment and AI. Or write to admin@wearetheartmakers.com

You provide
A Linux server with Docker (32 GB RAM recommended, 50 GB+ storage), optionally your own RPC endpoints and a local or cloud AI model.
We never ask for
Confidential datasets, API credentials, private keys or passwords — not by email and not during the pilot.
Terms
Duration, scope and support are agreed individually before a pilot starts.

Limitations

Stated plainly, so you know what the numbers cover.

Only the listed networks, verified protocol families and individually verified pools are analysed, and only pools that were discovered and ingested. Other venues are not counted, so no figure on this page is a token's whole market.
Pool-event volume counts a routed trade in every ingested pool it passes through. Transaction-net volume is reported separately.
The examples on this page cover historical, ingested windows. Nothing here is live, and the page connects to no API.
AI answer quality depends on the model. Validators reject ungrounded or forbidden statements but do not correct grammar.
Dune and cloud AI integrations are optional and off by default.
Off-site backups need an external S3-compatible destination configured per installation; until one is registered and verified, they are reported as not configured.
DeFi Intelligence is analytics and research software. It is not investment advice, makes no promise of trading or investment outcomes, and is not a regulatory certification.