Start here

The H-Series is a family of products that answer the questions an autonomous agent has to settle before anyone will trust it with real work: what can it find, what is it allowed to do, what did it actually do, and can that be checked afterward. Each product answers one of those questions, runs on its own, and composes with the rest into a single verifiable lifecycle.

The lifecycle

An agent doing real work moves through five stages. The H-Series has a product for each:

1. Discover what is out there (H-Index).
2. Get authority to act, scoped and capped (H-Grant, H-Agent, H-Cert).
3. Act, paying per action (x402 rails, H-Relay for transport).
4. Prove what happened (H-Seal).
5. Guard against misbehavior (H-Scope, H-Gate).

The products at a glance

• H-Index: discovery. A registry and semantic search over agents, tools, and MCP servers.
• H-Grant: authority + custody. Scoped, capped grants that let an agent use a credential it never sees (production nitro custody runs LIVE in a hardware enclave the operator cannot read, cutover 2026-07-26, so operator-zero-access is live; the software kms mode is retained on the standby as an instant rollback).
• H-Agent: agent identity. A runtime identity with an owner-signed on-chain spend cap; the agent holds an MPC key and spends only within its cap.
• H-Cert: standing. A directory of named principals with a behavior-backed standing tier, and owner-to-agent delegation.
• H-Seal: proof. Tamper-evident receipts of what an agent was asked to do and what it did, anchored on chain.
• H-Relay: transport. Mailbox, liveness heartbeat, and a context proxy so agents can reach each other.
• H-Scope: behavior. A posture score for an on-chain entity, pulled at the moment a trust decision is made.
• H-Gate: egress. Inspects what an agent is about to send and allows, redacts, or blocks it.
• H-Pact: membership. Owner-signed membership rings for group governance and access.

Three surfaces sit across all of it: the Console (drive an agent in plain language), SIEM (the estate’s audit and alerting view), and Reporting (activity and spend rollups).

The two front doors: showroom and agent

Every service has two ways in, and this Knowledge Base documents both:

• The showroom (the product’s .com site) is for a person: sign in, click through, see it work.
• The backend (the product’s .ai API) is for an agent or an integrator: signed requests, x402 payment, and where relevant an MCP tool surface.

Same product, same guarantees, two audiences.

Your first end-to-end path

A concrete walk through the lifecycle, using the Console as the easy on-ramp:

1. Open the Console and describe what you need. It searches H-Index for a service or connects to any MCP server you name, and shows you what it can do.
2. Create an agent and give it a budget. That budget is an owner-signed, on-chain H-Grant cap: the agent can spend, but only within it.
3. Run a tool. The agent acts under its grant and pays per call over x402. If the tool needs a stored credential, H-Grant releases it only inside the enclave; the agent uses it without ever holding it.
4. Check the receipt. Every paid run anchors an H-Seal receipt: proof of what happened that a third party can verify.
5. Stay guarded. H-Scope scores the counterparties involved and H-Gate can inspect what the agent sends, so a bad actor or a data leak is caught, not assumed away.

Where to go next

• Pick a service under Services in the sidebar for its full showroom and agent guide.
• New to running an agent? Start with the Console guide under Platform, or open the Console directly.
• Ready to give an agent its own identity? See agent onboarding.
• Want the architecture and trust model? See the architecture overview.