AiHummerAiHummer AiHummer
Sign in
Cloud or self-hosted · no Docker
v1.0.1 Releases & changes

The AI-employee platform for business — in the cloud or on your hardware, with no vendor lock-in

Multi-agent orchestration, channels for employees and customers, long-term memory, a plugin marketplace and an encrypted credentials vault. A finished product platform — not a library or an SDK. Run it in the AiHummer cloud or self-hosted on your own server.

Install → Documentation
Installation uses a personal link: sign up and get it in your account portal. Get it in your account →
orchestration always on
01
Sources
Telegram · MAX · Bitrix24 · WhatsApp · SIP · web · etc.
↓
02
Routing
smart routing by @mention
↓
03
Agent
an agent with memory, tools and permissions
↓
04
Delivery
guaranteed reply delivery
One lightweight server + PostgreSQL · deploys in minutes, no Docker
Principles Host-native No Docker No mandatory paid models Multi-tenant + SAML/LDAP/SCIM RLS Self-host observability
Capabilities

What's inside the platform

Everything you need to run AI employees in production — from a team of agents to guaranteed reply delivery. A finished product that runs on your hardware, with zero external SaaS.

01

Multi-agents

A whole team of AI agents with sub-agents: each with its own role, personality and model, with in-group routing by @mention.

02

Channels

Telegram, Bitrix24 (internal employee channel), SIP telephony, web widget, mobile client, generic webhook.

03

Memory

Agents that remember: Einstein pulls key facts from conversations, runs them through review, and keeps them with citations and a relation graph.

04

Knowledge · RAG

Upload your documents and knowledge base — the agent answers strictly to the point, citing its sources.

05

Marketplace

Extend the platform with one-click plugins — host-native, with auto-updates.

06

Credentials · vault

Every credential in an encrypted vault: personal and shared, with per-user OAuth2 connections.

07

Voice

STT → agent turn → TTS, diarization, translation and voice cloning via sidecars.

08

Control

Budgets and limits, guardrails/moderation, human-in-the-loop approvals, audit, RBAC, IP allowlist.

09

API

Embed AiHummer into any app through an OpenAI-compatible API and webhooks — with streaming, SCIM and device pairing.

10

Reliability

No reply ever gets lost: guaranteed delivery and conversation recovery after a restart.

Architecture

How it works

One lightweight server and PostgreSQL — deploys in minutes, no Docker. Messages flow into smart routing, then to the right agent with memory and tools, and the finished reply is guaranteed back to its channel.

Sources
Channels
Telegram · MAX
Bitrix24 · WhatsApp
SIP · web
etc.
→
Routing
by @mention
to the right
agent
→
Agent
memory & tools
model · persona
memory · perms
tools
→
Delivery
guaranteed
reply back
to channel
One server does it all
A single lightweight server is the admin console, the API and the conversation engine at once. The only hard dependency is PostgreSQL. It deploys in minutes.
Add capabilities as you grow
Search, a headless browser, STT/TTS, diarization, video, embeddings and a vector store — separate host-native services, enable only what you need.
Channels

Where your agents live

Reach employees and customers right where they already talk. Channels install from the marketplace in a couple of clicks — host-native, with shared credentials.

Telegram

Telegram

Bot channel for talking to employees and customers: text, media, mentions.

MAX

MAX

MAX, the Russian messenger: reach employees and customers in an app they already use.

Bitrix24

Bitrix24

Internal employee messenger — not customer omnichannel. Your AI employee replies right inside the Bitrix24 portal.

SIP

SIP telephony

Inbound and outbound calls via any carrier, post-call summary.

Web

Web widget

Embeddable site chat for customers and visitors.

iOS

Mobile client

Mobile and desktop app: QR-code sign-in, your company's branding and logo.

App StoreGoogle PlayAppGalleryRuStoreWindowsLinux
Webhook

Generic webhook

Connect email, cron and any of your own apps — accept arbitrary inbound events over a webhook.

And more channelssoon

WhatsApp Discord Slack Microsoft Teams Matrix Signal Email & more

WhatsApp, Discord, Slack, Teams, Matrix, Signal, Email — on the roadmap, coming to the marketplace.

Security

Security as the core, not an afterthought

Control and data ownership — on your own hardware. An encrypted vault, approval gates on risky tools, prompt-injection defense and enterprise auth in true self-host.

Air-gapped mode supported
Credentials vault
AES-256 encryption, personal & shared credentials
RBAC + RLS
roles and row-level security per tenant
Approval gates
human-in-the-loop on risky tools
Prompt-injection
defense and guardrails / moderation
Audit + IP allowlist
logging and network restrictions
SAML · LDAP · SCIM
enterprise auth + OIDC in self-host
Models & cost

No vendor lock-in on paid models

The working loop runs on free and local models. Want your own provider — drop in a key. A paid model is never required.

DEFAULT

Free / local

Free and local models + transport via a Codex/ChatGPT subscription. Zero mandatory token cost.

BYOK

Bring your own key

Each tenant supplies its own provider key. Billing and limits stay on your side.

OPENAI-COMPAT

Any provider

Anthropic, OpenAI and any OpenAI-API-standard endpoint: local models, OpenRouter, Gemini OpenAI-compat.

Memory

Einstein — our own memory that learns from every conversation

AiHummer's unique long-term memory architecture. Ships in the base install — works out of the box, locally, with no external paid services.

✓Semantic and episodic recall + structured fact extraction from every conversation
✓Background "dream" consolidation and ambient memory: accumulated context in every agent reply
✓A local embedder (multilingual-e5) — the agent's data and knowledge never leave your server
More about Einstein memory →
einstein · memory in the base install
01Conversation
↓
02Fact extraction
↓
03Memory + "dreams"
↓
04Context in every reply
Marketplace

Plugins — three tiers

From host-native channel services to zero-code wiring of any OpenAPI or MCP server.

TIER 1

Services: channels & voice

Separate repos, host-native, installed from the marketplace, shared creds: Telegram, Bitrix24, SIP, web widget, mobile client.

TIER 2

In-process skills

Personal integrations (Apple, Google, etc.) as Go packages inside the gateway, personal creds via credvault.

TIER 3

Zero-code sources

openapi — a tool per spec operation; mcp — connect any MCP server. No integration code.

Official plugins

Einstein

memory

Long-term memory: Markdown corpus indexing, cited search, embeddings, relation graph, review and audit.

Python · host-native

SIP Voice

voice

Real calls over any SIP carrier: inbound and outbound, hand-off of hard questions to the assistant, touch-tone dialing for IVR and call recording.

Python · baresip

Pocket Agent

Short-lived remote machine access via a one-time ticket: outbound connections, command allowlist, full audit.

Node/TS · Rust

Bitrix24

Bitrix24 portal as an internal employee messenger: text, media, mentions, reactions; REST pull model.

Node.js · stdlib

iOS / Mobile

Mobile and desktop chat app: real-time, files and push notifications, your company's branding.

iOS · macOS
View all plugins →
Your own plugin

Publish your own plugin

Build a plugin from a single manifest.json with the Plugin SDK — then publish it in one of two modes.

Private — for yourself

Side-load into your own instance from the admin UI. The signed artifact never leaves it.

Public — for everyone

A PR to the public registry: automated checks and review, then AiHummer counter-signs.

How to publish →

Official AiHummer plugins carry an Official badge; third-party community plugins connect as a separate source.

Deployment

One install.sh — and you're in prod

Host-native: a release tarball + systemd, everything under /home/.aihummer. No containers, podman, OCI or orchestrators.

Linux PostgreSQL · pgcrypto systemd
Get it in your account →

Don't want your own server? The same AI employees are available in the AiHummer cloud — a cloud plan deploys in a couple of clicks.

install — bash
# your personal link — in the my.aihummer.ru account portal
$ curl -fsSL "$INSTALL_URL" | sudo bash
→ unpacking to /home/.aihummer
→ systemd: aihummer-gateway.service
→ schema migrations applied on boot
→ server up and ready
✓ done · open the admin console
Comparison

Where AiHummer stands

A finished product versus toolkits. Runtimes, coding agents and frameworks are great for developers, but none of them is a ready-to-run AI-employee platform with channels, multi-tenancy and voice out of the box.

AiHummer compared with other agent systems across 13 capabilities. ✓ — yes, ~ — partial, — — no.
SystemTypeFinished self-host productWeb admin (no-code)Multi-tenant + RLSEnterprise SSO (SAML/LDAP/SCIM)RBAC + auditChannels: staff + customersVoice / telephonyLong-term memoryRAG / knowledge basePlugin marketplaceSecrets vault + approvalsNo Docker (host-native)
AiHummerProduct platformYesYesYesYesYesYesYesYesYesYesYesYes
ZeroClawAgent runtimeYesYesNoNoNoYesNoYesYesYesPartialNo
Hermes AgentAgent runtimePartialPartialNoNoNoPartialNoYesPartialNoNoPartial
PicoClawAgent runtime (lightweight)PartialNoNoNoNoPartialNoPartialNoNoNoYes
NanoBotAgent runtimePartialPartialNoNoNoPartialNoYesPartialNoNoPartial
MoltisAgent runtimeYesYesNoNoNoYesNoYesYesPartialNoNo
OpenHandsCoding agentYesYesNoNoNoNoNoYesYesPartialPartialNo
AiderCoding agent (CLI)PartialNoNoNoNoNoNoYesYesNoNoYes
ClineCoding agent (IDE)YesYesNoNoNoNoNoYesYesYesPartialYes
ContinueCoding agent (IDE)YesYesNoNoNoNoNoYesYesYesNoYes
GooseCoding agentYesPartialNoNoNoNoNoYesYesYesPartialYes
LangGraphFramework / SDKNoYesNoNoNoNoNoYesYesPartialPartialYes
CrewAIFramework / SDKNoPartialNoNoNoNoNoYesYesYesNoYes
AutoGenFramework / SDKNoYesNoNoNoNoNoYesYesPartialNoYes
Google ADKFramework / SDKNoPartialNoNoNoNoNoYesYesNoNoYes
OpenAI Agents SDKFramework / SDKNoNoNoNoNoNoPartialYesYesNoNoYes
SmolagentsFramework / SDKNoNoNoNoNoNoNoPartialYesNoNoYes
Semantic KernelFramework / SDKNoNoNoNoNoNoNoYesYesPartialNoYes
MastraFramework / SDK (TS)NoYesNoNoNoNoPartialYesYesPartialNoYes
HaystackRAG frameworkNoPartialNoNoNoNoNoPartialYesYesNoYes
LlamaIndexRAG frameworkNoPartialNoNoNoNoNoYesYesYesNoYes
DifySelf-host product (closest)YesYesYesYesYesYesNoYesYesYesPartialNo

A category-level capability comparison: a finished product versus runtimes, coding agents and frameworks, which solve different problems.

Why AiHummer

Why companies choose AiHummer

Your data, your control

Runs on your own hardware: conversations, credentials and knowledge never leave your perimeter — covering data-protection and security needs.

No vendor lock-in, no per-token bills

Free and local models or BYOK. A fixed license, not a bill for every message.

One product instead of a SaaS zoo

Channels, long-term memory, RAG, voice, marketplace and an admin UI in one platform. A finished product, not a library.

Live in minutes, no Docker

One light server plus PostgreSQL, installed with one command. Host-native, no container hassle.

Enterprise-ready

Multi-tenancy, SAML/LDAP/SCIM, Row-Level Security, audit and human-in-the-loop approvals — out of the box.

Built for the Russian market

MAX and Bitrix24, SIP telephony, import substitution and the Russian software registry — things foreign platforms lack.

Results in numbers

What an AI employee saves

E-commerce$54 684saved a yearMedical clinic$49 997saved a yearReal estate$140 616saved a yearOnline school$54 684saved a yearInternal helpdesk$89 057saved a yearRealtor$1 440protected income/yr

Model estimates on transparent assumptions — see the calculator.

Case studies

How AiHummer pays for itself

Typical industry scenarios — not real named clients. Every figure reproduces from the calculator assumptions.

E-commerce

Online store, ~300 requests a day

Before

Two operators are swamped at peak and offline at night — order status, delivery and returns pile up.

After

An AI employee closes ~65% of routine requests in Telegram and MAX around the clock, escalating only edge cases.

$54 684Saving/yr
13 daysPayback
65%Automated

How it works: Telegram + MAX channels, a support agent, knowledge-base RAG, an order-status plugin.

Model: −1 operator FTE (₽55,000/mo), full cost ×1.302. 24/7 cover is on top of the saving.

Calculate for yourself →
Medical clinic

Clinic: receptionists can’t keep up

Before

Calls and messages get lost, bookings slip, patients leave for competitors.

After

An AI receptionist handles booking, rescheduling, reminders and visit prep in messengers 24/7.

$49 997Saving/yr
5 daysPayback
60%Automated

How it works: Telegram/WhatsApp channels, a receptionist agent, schedule integration, reminders.

Model: −1 receptionist FTE (₽52,000/mo). Market anchor: up to ~70% of bookings via messenger.

Calculate for yourself →
Real estate

Developer: leads arrive at night and on weekends

Before

Site, Avito and Telegram leads wait until morning — up to 40% arrive off-hours.

After

An AI employee qualifies a lead in a minute at any hour, books a viewing and hands the manager a warm contact.

$140 616Saving/yr
5 daysPayback
45%Automated

How it works: Site widget / Avito / Telegram channels, a qualification agent, calendar booking, a CRM plugin.

Model: −1.2 sales FTE (₽90,000/mo). Reply ≤1h → ~7× higher qualification odds (HBR).

Calculate for yourself →
Online school

Online school: student questions drown in chats

Before

Tutors spend the day on routine questions about the syllabus, access and instalments.

After

An AI curator handles ~60% of questions and helps sell courses, freeing tutors to teach.

$54 684Saving/yr
5 daysPayback
60%Automated

How it works: Telegram/web channels, a curator agent, syllabus RAG, a payments/access plugin.

Model: −1 tutor FTE (₽55,000/mo). Anchor: “1 AI assistant per 10 tutors” (BotHelp).

Calculate for yourself →
Internal helpdesk

B2B company: first-line IT/HR in Bitrix24

Before

IT and HR specialists drown in repetitive tickets: access, how-tos, statuses.

After

A first-line AI employee in Bitrix24 answers staff instantly and files tickets itself.

$89 057Saving/yr
8 daysPayback
55%Automated

How it works: Bitrix24 channel, a helpdesk agent, policy RAG, a ticketing plugin.

Model: −1.5 specialist FTE (₽60,000/mo). Anchor: “up to ₽5M/yr per 10 staff” (AutoFAQ).

Calculate for yourself →
For one person
Realtor

Realtor: listing enquiries 24/7

Before

Listing calls and messages arrive at night and in busy hours — 17–21% are missed.

After

A personal AI assistant answers on Avito and Telegram, qualifies and books viewings to the calendar.

$1 440Protected/yr
answered in seconds, the lead stays warm 

Model: ₽50,000/mo ads at ₽2,000/lead = 25 leads; ~20% lost off-hours ≈ 5 leads/mo. Deal commission ₽100–300k — one saved deal pays back a year of licence many times over.

Calculate for yourself →
Private tutor

Private tutor / expert

Before

Requests come in the evening, some are lost, and messaging eats teaching time.

After

An AI assistant takes requests, manages the schedule, reminds students and answers routine questions.

$6 720Protected/yr
no request lost, instant reply 

Model: ~3–4 of ~20 monthly requests lost to evening silence × 8 lessons × ₽1,500 ≈ ~₽42,000/mo protected income.

Calculate for yourself →
Avito seller

Seller on Avito / Telegram

Before

“Still available?” arrives any time; a slow reply loses the buyer to the next seller and drops your ranking.

After

An AI employee replies in seconds any time, closes the deal and calls a human for edge cases.

Free startCommunity
answered in seconds, 24/7 

Avito recommends replying within 10–12 min, response time is public on your card, misses lower your ranking. Free start on Community.

Calculate for yourself →
Industries

Who buys AI employees

90% of bank deployments are on-premises
  • Customer support and requests in messengers
  • Product/status answers with no PII leakage
  • Data stays in your perimeter
Fixed price

Why a flat licence beats per-dialog pricing

Cloud bots charge per dialog, per resolution or per token — your bill grows with your traffic. AiHummer is a flat licence: load grows, the bill does not.

Load grows — the bill does not
Your data, your hardware

Self-hosted, no telemetry

  • Data never leaves your server — conversations, credentials and knowledge stay in your perimeter.
  • 152-FZ: since 01.07.2025 primary PII collection must be in Russia; repeat-breach turnover fines of 1–3% of revenue, up to ₽500M.
  • 90% of bank deployments are on-premises. No telemetry, no back-channel to the vendor.
ROI calculator

Estimate your saving

Saving ≈—/yr
Full calculation →
Referral program

Bring a friend — get +10 days

Share your personal link from your AiHummer account. When a friend registers through it and activates a paid subscription, you get +10 days added to any paid plan. No limit — every subscribed friend is another +10 days.

+10 days per friend
01Take your link in the portal

Your personal referral link lives in your account at my.aihummer.ru — copy it in one click.

02A friend registers & subscribes

They sign up through your link and activate any paid subscription — attribution is automatic.

03You get +10 days

+10 days are added to any of your paid plans. There is no cap on how many friends you invite.

Open your account →

The reward is credited once your invited friend activates their first paid subscription.

Blog

From the AiHummer blog

Product updates, guides and notes on AI employees — cloud and self-hosted.

All articles →
1 min read

Привет! Я Hum — маскот AiHummer, тот самый лисёнок в фирменном ошейнике 🦊

Это официальный канал aihummer.ru. Здесь я буду общаться с вами, рассказывать про новые фишки и релизы нашей платформы ИИ-сотрудников, а ещё делиться самым…

Pricing

Simple, predictable pricing

Community is free forever. Self-hosted is a fixed license with no per-message billing; cloud plans give you a ready server in the AiHummer cloud.

Community

Free

Self-host, 1 tenant, up to 2 AI employees, core channels.

Starter

from $58/mo
billed yearly · $69/mo monthly

Commercial use, up to 3 AI employees, all channels.

Business

from $166/mo
billed yearly · $199/mo monthly

White-label, RBAC, audit, up to 5 AI employees.

Enterprise

On request

Multi-tenant, SAML/LDAP/SCIM, RLS, audit, SLA, RU registry.

See full pricing →
About us

The AiHummer team and Hum the fox

AiHummer is a Russian product team. We build an AI-employee platform you can deploy on your own hardware or run in the cloud: channels, memory, a marketplace and security — one finished product, not a construction kit.

More about the company →

Hum is our mascot — a fox cub wearing the AiHummer collar. He greets you on the site, in the app and in the docs — a reminder that an AI employee should be quick, friendly and always around.

Deploy AI employees in minutes

In the AiHummer cloud or on your hardware — no Docker, no vendor lock-in. Your data stays with you.

Install → Documentation