AiHummer× HummerTech

AiHummerWorking product · Native AI layer

AI employees
that get things done.

Agents use knowledge and memory, work in every channel, take action through real connections and hand complex cases to a human. AiHummer is the company's first product and the native agent runtime of the whole ecosystem.

Open AiHummer View the documentation

AvailableUsers and organizations of any scaleSaaS & self-hosted

The AiHummer Gateway interface: request routing and connections — The real AiHummer Gateway interface · the product is available at aihummer.ru

Where assistants stop

The assistant replies. A human still does the work

Most AI assistants end at text. Between the answer and the result, manual work remains.

An answer without action

The model states what should be done, but a human has to do it in the CRM, the bank or the e-document system.

Memory does not outlive the chat

Every new session starts from scratch: customer context, agreements and decisions are lost.

Knowledge apart from processes

The knowledge base lives in one place, while data and actions live in other systems.

Unclear what the agent may do

Without explicit policies and confirmations, autonomy turns into risk.

Native depth

Native AiHummer and compatible providers

Integration depth differs — and the page says so plainly.

Native

AiHummer in AiHummer

AiHummer is the native AI layer of HummerTech. What other products get as 'the AiHummer role' is the product itself here: agents, memory, knowledge, tools and confirmations.

The deepest access to Hummer Context
Policies and autonomy levels come from Hummer Trust
Every action leaves a trail in Hummer Proof
Complex scenarios are built in Hummer Flow

Open AiHummer Role in the ecosystem

Compatible

Third-party agents and models

Third-party models and providers connect through the AI Gateway: routing, limits and model choice per task. A third-party agent can work alongside, with access to context and actions governed by separate policies.

Routing goes through the AI Gateway, permissions through Hummer Trust. Setup takes extra work, and access to context and actions is limited by the organization's policies.

Connect your own agent

An agent with permissions

An agent with memory, permissions and tools

AiHummer combines Einstein memory, corporate knowledge, multi-agent orchestration and tool calling on top of real connections — with a mandatory human handoff where required.

Einstein Memory. native memory of dialogs and decisions, isolated per context
RAG Knowledge. answers based on the organization's permitted documents
Multi-agent. several roles solve a task together and hand it to each other
Tool-calling. calling tools and actions in the source systems
Human-in-the-loop. handoff to a human with context preserved and a proposed solution
Analytics. review of dialogs, request causes and agent quality

What agents do

What agents do today

AiHummer is a working product. Detailed scenarios, pricing and documentation are on the product site.

Reply in the customer's channels

Text and voice requests are handled in the channels the customer already uses, with one dialog history.

Take action

Through tools and connections the agent creates a record, sends a document or starts a check — in the source system.

Hand off to a human

When confidence is low or confirmation is required, the agent passes the task to an operator with the context and a draft solution.

Platform composition

AiHummer modules. Detailed descriptions and current capabilities are in the product documentation.

Module

Agents

Roles, instructions and autonomy boundaries.

Module

Instructions

Behavior rules and answer style.

Module

Einstein Memory

Long-term memory per context.

Module

RAG Knowledge

Search across permitted documents.

Module

Channels

Text and voice request channels.

Module

Tools

Calling tools and actions.

Module

Plugins

Platform extensions.

Module

Multi-agent

Orchestration of several agents.

Module

Human-in-the-loop

Human handoff and confirmations.

Module

Analytics

Quality, causes and outcomes.

Channels and tools

The source of truth stays with you

AiHummer uses the same connections as the rest of the ecosystem. Data is read as permitted read models, and actions run in the source systems.

Native depthShared Context, Trust, Proof

Every layer rests on HummerFabric: connections, context and policies are not duplicated inside the product.

MessengersTelephonyMailCRM1CKnowledge basesInternal APIsMCP servers

Integration catalog and statuses

Autonomy

Agent boundaries, memory and deployment

Identity, connections, permissions, confirmations and deployment — the ecosystem's shared mechanisms applied to this product.

HummerAuth

Contexts in AiHummer

AiHummer works in the context the user is in. Personal memory does not enter the corporate perimeter, and corporate memory does not reach personal devices outside policy.

PersonalProfessionalCorporatePortal (external parties)

How HummerID works

Connection Vault

Connect once

The agent's tools rest on connections from the Connection Vault. The agent's access can only grow by widening the connection itself — and the owner sees it.

The owner sees the list of connections, scopes and terms, can narrow the data volume and revoke access in one click.

Connection Vault in Fabric