Questions, answered. Everything about Inflectiv.

Inflectiv turns unstructured information — documents, PDFs, spreadsheets, research, institutional know-how — into structured datasets that AI agents query directly, over API, SDK or MCP. The agent pays per question. The owner sets the terms and earns on the questions it is asked.

The part that makes them living rather than static: agents can write back what they learn, so a dataset improves while it is being used instead of going out of date the moment it is delivered.

Inflectiv serves four core user groups:

• Data contributors and creators: researchers, domain experts, professionals, and institutions sitting on valuable knowledge they've never been able to monetize.
• AI developers and labs: teams that need clean, domain-specific, provenance-aware data their agents can query at decision time.
• Enterprises: companies looking to structure internal data for AI workflows without giving up ownership or control.
• Builders and AI agents: projects building agentic applications that need a queryable, provenance-aware source they did not have to build a pipeline for.

The flow is end-to-end:

1. Ingest: Upload files, point it at URLs, or type knowledge straight in. Inflectiv structures the material into records, embedding and de-duplicating them as they land.
2. Encrypt and store: Records are encrypted with Seal (Mysten Labs' decentralized secrets management), the access policy you set is held on chain, and the data is stored durably on Walrus.
3. Set the terms: Public or private, free or paid, and who — if anyone — is allowed to write to it.
4. Query: The dataset is queryable the moment it exists, by agents, apps, APIs and MCP workflows, with no pipeline in between.

Tokenizing is a fifth step, and an optional one. A dataset works, earns and is queried whether or not it ever has a token.

Three ways, and most agents use more than one:

• MCP: Inflectiv exposes twenty-seven tools over the Model Context Protocol. The server has been live since April and five clients reach it, including Claude and Cursor, so an agent connects with one line and no connector to build.
• API and SDK: Authenticate, satisfy the dataset's access policy, and query it as part of the reasoning loop. Ask for an answer and you get one back with the records it used — each with its source, its date and a similarity score. Ask for the records alone and you get them ranked, with nothing generated, so you can ground your own model instead.
• Write-back: Turn on the attachment mode and an agent adds what it learned back into the dataset as it works.

A dataset also carries a 5,000 character instruction field. Whatever you write there lands in the prompt of every agent attached to it, so you can change how your data gets used without touching anyone's code.

One credit per question. Not per row, per result, per content type or per effort tier — a question costs a credit, which means you can work out a bill without a spreadsheet. Asking for the ranked records costs the same as asking for the answer, so the choice is about your architecture rather than your bill.

Every query carries a score threshold. If nothing clears it you get nothing back and you are not billed, so a bad question is cheap to ask.

You buy credits on a plan or as a top-up, and there is a free tier to start on. When credits run out an agent can settle in USDC over x402 on Base and carry on — no card, no invoice, and nobody approving anything.

Anything that can become structured intelligence. Today the platform hosts datasets across:

• Research papers and academic knowledge
• Financial and on-chain market signals
• Industry-specific know-how (legal, medical, engineering, etc.)
• Spatial and geospatial data
• AI agent activity logs and behavior traces
• Enterprise internal knowledge bases

Traditional data marketplaces sell static files. Once you pay, you download, and the relationship ends. Inflectiv works differently:

• Ownership stays with the creator. Buyers get access, not ownership. They get answers, not your files.
• It is priced by the question. A dataset is not a one-off sale — the one that gets asked the most is the one that earns the most, for as long as it keeps being asked.
• It improves after you buy it. With write-back on, the agents querying a dataset leave it better than they found it.
• Provenance travels with the answer. Every dataset and record is on chain, on Base or Sui, and every answer hands back the records behind it with their sources, dates and scores.
• Built for agents, not just humans. Datasets are queryable at decision time, not downloadable as files.

$INAI is the protocol's settlement layer rather than a consumer-facing utility token. Its role is infrastructural: every economically meaningful action inside Inflectiv eventually resolves into a settlement event denominated in $INAI — dataset token creation, bonding curve participation, liquidity provisioning, and buybacks triggered by usage.

• Fixed supply: 16 billion tokens, with no emissions tied to activity, usage or participation. Growth consumes supply rather than expanding it.
• Usage-driven buybacks: When a dataset's cumulative paid usage crosses defined thresholds, a portion of protocol revenue buys back the corresponding dataset token in $INAI, and those tokens are permanently burned.
• One accounting unit: Centralizing settlement here keeps value flows observable and auditable, instead of leaking into parallel reward tokens or isolated incentive pools.

Developers do not need to hold or understand $INAI to build on Inflectiv. Access runs on credits; $INAI operates beneath that abstraction.

A creator can choose to launch a dataset with its own token. That token is a market hypothesis about how useful the dataset will turn out to be — a way for participants to express conviction through price discovery. It is not an access key: holding one does not gate APIs, unlock features, or substitute for credits.

The creator retains up to 49 percent of supply, locked for 30 days and then unlocking linearly, so there is no exit before the dataset has shown real adoption. The token then enters a bonding curve denominated in $INAI, initialized with virtual reserves so no upfront capital is required, and the curve prices it algorithmically as demand arrives.

The curve stays open until it reaches a maturity threshold — currently around $5,000 USD-equivalent, and deliberately adjustable. At that point it closes, the token graduates to a DEX (Momentum on Sui or Aerodrome on Base), and the liquidity is time-locked. Graduation is a transition, not an endpoint: from there the token's value is increasingly coupled to actual dataset usage rather than speculative inflows.

Two ways, and it is worth being precise about which is which:

• Access fees: You set the terms on your own dataset — public or private, free or paid, and at what price. A dataset that keeps being queried keeps earning, which is the opposite of selling a file once.
• Token appreciation: If you tokenized the dataset, your retained allocation appreciates as genuine demand for it grows.

There are deliberately no per-query token royalties. Paid usage instead triggers protocol-level buybacks that burn the dataset token, and only paid SDK usage counts — free queries, internal testing and self-referential calls do not. That is what stops anyone inflating usage to extract value, and it is why the datasets that do well are the ones that are genuinely useful rather than the ones marketed hardest.

Sui's architecture is uniquely suited to a tokenized data marketplace: object-centric storage, fast finality, low fees, and zkLogin for frictionless wallet onboarding. Combined with Walrus for decentralized storage and Seal for access control, the Sui stack gives Inflectiv durability, speed, and cryptographic access enforcement in one tightly integrated package.

Inflectiv is already multichain. Datasets and records are written on Base or Sui, agents settle in USDC over x402 on Base, and dataset tokens graduate to a DEX on either chain. $INAI keeps a single canonical supply across all of them: tokens are minted on a source chain and bridged using standardized lock-and-mint mechanisms, so bridging never creates supply or fragments liquidity.

You do. Uploading to Inflectiv does not transfer your intellectual property — we structure, store and serve your data, and you keep control of it.

That is enforced rather than promised: Seal encrypts the records, the access policy you set lives on chain, and buyers receive answers and ranked records, never your source files.

There are three on-ramps depending on what you're trying to do:

• Contribute a dataset: Login with your wallet or with your Google account, upload your files, and follow the structuring flow. A free tier is available to start.
• Build with the API or MCP: Point your agent at Inflectiv and start asking. Documentation lives at docs.inflectiv.ai.
• Browse the marketplace: Explore live datasets and query the ones your agents or systems need.

For partnerships, enterprise pilots, or anything not covered here, email [email protected].