Private beta, Q4 2026. Join the waitlist

On-demand blockchain data

Ask any question about any EVM chain over CLI or MCP. Testril returns the answer, served from cache or computed on demand, with every value traceable back to its source blocks.

All EVM chains No schemas Any question

* Early members get free query credits at launch.

You're on the list. We'll email you when your access opens.
01 / Ask

State what you want

Say it in plain language over CLI or MCP. No schema, no subgraph, no pipeline to stand up. If it happened on chain, ask for it.

02 / Get a quote

See the price before anything runs

Testril prices the request up front. You see what the data costs and what it takes to produce, before a single block is touched.

03 / Accept and pay

Review the quote. Pay. Receive your data

We support x402 and MPP. Your data will be encoded in self-describing JSON.

04 / Verify

Wonder where the data came from?

Ask for provenance. We'll tell you where the data came from and how we computed it.

ai-agent · terminal
testril · data runtime
endpoints · active

Provenance attached to every endpoint: source events, block range, transform, and latest indexed block.

The problem

Agents don't need raw blockchain data

Agent pays too much

Raw blocks, logs, and traces create a huge context. Agents pay to ingest and then to reason over all that data.

APIs lack data coverage

If your question fits an existing endpoint, you're fine. If it doesn't, you're stuck. APIs were built for humans.

Data cannot be audited

Agents have to trust the data they work with. Without provenance they can't verify anything.

93–99%*

fewer inference tokens per answer than an agent crunching raw data.

100+

EVM chains at beta, so new questions don't stop at chain boundaries.

100%

of values traceable to source data and the code that produced them

* Range across internal benchmark queries: a Testril endpoint versus an agent computing the same result over raw indexed tables through an MCP.

The architecture

Compute and store only what's needed

Index everything

The traditional model

Compute and store every block on every chain up front, expose the tables, and leave interpretation to whoever asks.

  • Compute and storage used for data no ever asks for
  • Schemas are fixed at indexing time, so new questions need a new pipelines
  • Interpretation falls to the agent, at inference prices
  • Derivative data isn't auditable

Index on demand

The Testril model

The question arrives first. Testril reads and store only the data it needs. Transformations are done once, and the results cached.

  • No data is indexed or stored until it's needed
  • Warehouses aren't designed to answer questions someone might ask
  • Agents receive answers, not piles of data to analyze
  • Data lineage and transform code ship with every response

What becomes possible

If it happened on chain, your agent can ask about it

Stablecoins

Flow analysis

Compare USDC and USDT volume across any block range and keep the numbers current.

Liquidity

DEX monitoring

Track pool changes, net liquidity, asset exposure, and protocol activity as they happen.

Wallets

User cohorts

Find wallets by behavior, value threshold, bridge activity, or what they touched next.

Risk

Exposure checks

See which wallets or protocols touched a given asset, pool, bridge, or contract.

Reporting

Live metrics

Feed current, verifiable numbers into products, dashboards, and internal reports.

Research

The unexpected question

Ask the thing nobody built a dashboard for. If the data is on chain, it's in reach.

Who it's for

Built for agents. Usable by developers

AI agent teams

Give agents live and historical chain data without loading raw data into every request.

agentstoolsworkflows

dApps and protocols

Power custom metrics, alerts, liquidity views, and reporting from a single endpoint.

appsanalyticsmonitoring

Financial products

Use current, verifiable chain data for risk, research, trading, and reporting.

riskresearchreporting

FAQ

Common questions

Is this an MCP server?

No. Testril is a data network. MCP is one of the interfaces into the network. We also provide a CLI.

Is this another blockchain indexer?

Yes, in a sense. Traditional indexers prepare datasets before anyone asks. Testril starts with the question and computes on demand, so nothing is computed or stored until someone asks for it.

Why not just point an MCP at indexed data?

Three reasons:

1. Using raw data is expensive!

2. Raw data is pre-organized, but maybe not the way you need it.

3. Raw data isn't auditable.

How does pricing work?

You get a quote before anything runs. Pay per request for a one-off question, or stand up a dedicated indexer when you need the data continuously. No subscription required to ask.

Can I get real-time data?

Yes. For data you need on an ongoing basis, Testril provisions a dedicated indexer that keeps feeding it to you as new blocks finalize.

How do I know the data isn't hallucinated?

Every answer is auditable: both the provenance of the raw data and any transformation code that produced it.

Which chains are supported?

All EVM chains at beta. Support for additional ecosystems will follow based on demand from early members.

When can developers use it directly?

Developers can use it now.

Your agents have questions. Give them endpoints

* Early members get free query credits at launch.

You're on the list. We'll email you when your access opens.