# Inflectiv

> Inflectiv turns unstructured information into structured datasets that AI
> agents query directly, over API, SDK or MCP. The agent pays per question, the
> owner sets the terms, and agents can write back what they learn — so a dataset
> improves while it is being used instead of going stale the moment it ships.

Written for AI systems reading this site. Everything below is present tense
because it has shipped; anything still to come is marked as such.


## How an agent reaches a dataset

- MCP: twenty-seven tools over the Model Context Protocol. The server has been
  live since April and is reachable from five clients, including Claude and
  Cursor. One line to connect, no connector to build.
- API and SDK: authenticate, satisfy the dataset's access policy, and query it
  inside the reasoning loop. Ask for a cited answer and you get the records
  behind it — each with its source, its date and a similarity score. Ask for the
  ranked records alone and nothing is generated, so you can ground your own model.
- Write-back: with attachment mode on, an agent adds what it learned back into
  the dataset as it works.
- Instructions: every dataset carries a 5,000-character instruction field that
  lands in the prompt of every agent attached to it.


## What a question costs

- One credit per question. Not per row, per result, per content type or per
  effort tier. Asking for ranked records costs the same as asking for an answer.
- Every query carries a score threshold. If nothing clears it, you get nothing
  back and you are not billed.
- Credits come with a plan or as a top-up, and there is a free tier.
- When credits run out an agent settles in USDC over x402 on Base and carries
  on — no card, no invoice, no human approving anything.


## How a dataset is built and held

- Ingest files, URLs, documents, research or typed-in knowledge; Inflectiv
  structures it into records, embedding and de-duplicating as they land.
- Records are encrypted with Seal, the access policy is held on chain, and the
  data is stored on Walrus.
- The owner sets the terms: public or private, free or paid, and who may write.
- Datasets and records are written on Base or Sui.


## The economy

- $INAI is the settlement layer, not a consumer utility token. Fixed supply of
  16 billion, with no emissions tied to activity, usage or participation.
- A dataset may optionally launch its own token. That token is a market
  hypothesis about the dataset's usefulness. It is NOT an access key: holding
  one does not gate APIs, unlock features or substitute for credits.
- Creators earn two ways only: access fees they set, and appreciation of tokens
  they retained. There are deliberately no per-query royalties.
- Paid usage triggers protocol-level buybacks that burn the dataset token. Only
  paid SDK usage counts, so usage cannot be inflated to extract value.
- Proof of Knowledge and DataLabs are the model that extends this: domain
  economies competing each epoch for a known pool of INAI, with contributors
  earning by the proven value of what they added.


## Pages

- [Home](https://www.inflectiv.ai/): what living datasets are and how the loop closes.
- [MCP Server](https://www.inflectiv.ai/mcp): the twenty-seven tools, the clients, the install line.
- [Compare](https://www.inflectiv.ai/compare): Inflectiv against search APIs, licensed corpora and building your own — including where the alternatives are better.
- [Use Cases](https://www.inflectiv.ai/usecases): six apps built on Inflectiv and the architecture they share.
- [FAQs](https://www.inflectiv.ai/faqs): thirteen questions answered plainly.
- [Proof of Knowledge](https://www.inflectiv.ai/proof-of-knowledge): the DataLab economy.
- [INAI and DataLab Tokens](https://www.inflectiv.ai/inai): what each asset does.
- [Tokenization](https://www.inflectiv.ai/tokenization): bonding curves and usage-driven buybacks.
- [Blog](https://www.inflectiv.ai/blog): releases, architecture notes and industry writing.
- [Press Kit](https://www.inflectiv.ai/press-kit): logos, palette and media resources.


## Who it is for

- AI developers and labs needing clean, domain-specific, provenance-aware data
  their agents can query at decision time.
- Enterprises structuring internal data for AI workflows without giving up
  ownership or control.
- Researchers, domain experts and institutions with knowledge they have never
  been able to monetize.
- Web3 builders and agent developers who would rather query a dataset than build
  a pipeline.


## Ecosystem

Sui, Walrus, Seal, Mobula, ASI, DogeOS, ElizaOS, Codec.


## Elsewhere

- App: https://app.inflectiv.ai
- Pricing: https://app.inflectiv.ai/pricing
- Docs: https://docs.inflectiv.ai/docs/overview
- X: https://x.com/inflectivAI
- LinkedIn: https://www.linkedin.com/company/inflectivai/
- Support: support@inflectiv.ai
- Press: press@inflectiv.ai
- Sitemap: https://www.inflectiv.ai/sitemap.xml