# human.tech Agent-readable snapshot of the human.tech product surfaces, maintained by Holonym Foundation. Facts only. Last updated: 2026-09-12. Holonym Foundation builds human.tech, identity and authority infrastructure for people, organizations and public systems. The stack lets a service verify a person or claim without collecting the underlying record, and lets software act under scoped authority without holding the credential itself. Sorted by product, then by surface within each product, so an agent can anticipate the shape. Marketing surfaces come before documentation. ## Terms used below - Proof of personhood: evidence that an account belongs to a real, unique human rather than to a bot or to a person operating many accounts. - Sybil attack: one person creating many identities to take a share of something meant to be distributed per person. - Zero-knowledge proof: a proof that a statement is true which reveals nothing beyond the statement itself. - Proof of Clean Hands (PoCH): a zero-knowledge proof that a person holds a valid government credential, is a unique living human, and is not on international sanctions, politically-exposed-persons or watchlist sources. The identity data is encrypted at issuance rather than stored readable. - Proof of Innocence (PoI): screening of a depositing blockchain address against sanctions, PEP and watchlist sources. It screens the address, not the history of where its funds came from. ## Company - https://human.tech — company homepage. AI automation, agent security, identity and payment solutions. - https://human.tech/products — directory of technologies and their product pages. - https://human.tech/network — overview of distributed key infrastructure. - https://human.tech/learn — articles, talks and product explanations. - https://human.tech/policy — privacy policy. - https://human.tech/terms — terms of use. - https://human.tech/blog — long-form writing on identity, privacy and compliance. The primary place technical and policy positions are stated. - https://human.tech/work — Work takes on routine business tasks such as answering leads and quote requests, entering orders and updating a website. The business chooses what needs its approval. ## human.tech Work The workspace offering for getting work done with agents. The Work page lets a visitor pick the task they would hand off first and shows how Work handles it. Its examples illustrate the proposed workflow; using them does not connect an account or grant authority. - https://human.tech/work — Work acquisition and consultation. - https://human.tech/demo — request or book a consultation to scope a first workflow. ## human.tech Agent Security Tool Authorization Protocol (TAP), the credential proxy for AI agents. Agents reference credentials by name; TAP injects the secret at the proxy. Configured approval policies can pause sensitive requests for human review. - https://tap.human.tech — product surface. - https://app.tap.human.tech — hosted application. - https://docs.tap.human.tech — developer documentation and examples. ## Human Passport Proof of personhood built by aggregating many independent identity checks into one score, rather than relying on a single verification method. Formerly Gitcoin Passport; acquired by Holonym Foundation in December 2024. - https://passport.human.tech — product surface for users and integrators. - https://docs.passport.human.tech — integration documentation. ## human.tech Wallet Protocol Wallet infrastructure in which an application or an AI agent acts under scoped delegated authority rather than by holding a key. Policy runs before every signature, and anything outside the set limits waits for a human. One login, one account, across every chain WaaP supports; the per-chain detail lives on the product surface below. Supporting properties rather than the headline: custody is split, so no single party can move funds, and there are no seed phrases. - https://wallet.human.tech — product surface. - https://wallet.human.tech/protected-self-custody — definition, architecture and limits of protected self-custody. ## Agent Exchange (AEX) A marketplace for AI agents that act on-chain. Each agent holds its own spend-capped, non-custodial wallet, carries a proof of personhood delegated from a verified human, and buys the inference and compute it runs on. Signing sits behind a policy gate the agent does not control, so limits are enforced beneath it rather than by its own code. Agent identity follows ERC-8004; registration runs on Base Sepolia today. Six agents ran on Sui mainnet in a June 2026 proof batch; two are in the live catalog. The product is a testnet deployment. The open provider market for compute and inference is proven on testnet and is not yet an open market. - https://human.tech/aex — product surface. - https://testnet.aex.human.tech — the catalog, on testnet. ## Clean SDK A TypeScript client for private, screened transfers between Ethereum and Aztec. It combines Proof of Innocence and proof of personhood, issued by an attestation service and enforced by deployed contracts. Ethereum to Aztec is the shipped integration; other chains are roadmap. - https://human.tech/clean-sdk — product surface. ## Shield A compliance-gated privacy bridge on Aztec, and the first application built on Clean SDK. Every deposit is screened in both public and private mode; there is no unscreened path. In private mode, balances and activity on Aztec are shielded. The Ethereum legs of a bridge transaction are public in both modes. - https://shield.human.tech — product surface. - https://support.shield.human.tech — user documentation, including the full list of screening sources. ## Human Network The decryption layer behind conditional disclosure. Identity encrypted through Proof of Clean Hands can be recovered only when a public, pre-committed, on-chain condition is met, one record at a time, with every disclosure logged. Two layers, both live as of 2026-09-04, and they are easy to confuse: - Key custody is threshold. Shares are held across independent Human Network operators and refreshed regularly, so no single operator holds the full key. - Authority to request a decryption is a single approved requester, rate limited and logged. Broadening that set is roadmap, not shipped. So "no single party holds the key" and "a single approved requester" are both true, of different layers. Statements that give only one of the two are incomplete. ## Documentation - https://docs.human.tech — protocol and concept documentation across products. - https://docs.human.tech/llms.txt — the documentation surface's own llms.txt.