Kormaitech

AI Agent Development

Your AI team, working where you already work

Professional AI agents for small and medium businesses. They manage tasks, answer on WhatsApp and Telegram, handle email, remind you of deadlines, watch your market, track your bills and draft your contracts.

Built on open-source engines: OpenCode, OpenClaw and Hermes. Secured, monitored and maintained by us.

Why start now

The tools are ready. Most businesses are not.

61%

of SMEs use at least one AI application

Only 3.6% deploy real agents that carry out multi-step work. Source: OECD, 2026.

<10%

of workers can define an AI agent

Fewer than 10% of workers can correctly define what an AI agent is. Source: Section AI Proficiency Report, July 2026.

That gap is your opening. One well-built agent on one real task beats ten tools nobody uses.

What your agent can do

Every agent is configured for your business, your tools and your language. These are the features businesses ask for first.

Team task management

Your team assigns work by message: “Ask Priya to send the quote by Friday.” The agent records the task, follows up with the owner, flags what is overdue and sends a clear status every morning. On the Small Team plan, each person gets a dedicated agent and the team shares one more.

WhatsApp and Telegram access

Talk to your agent like a colleague. Send a text, a voice note, a photo of an invoice or a PDF. It answers in the same chat, day or night. No new app to learn, no password to remember.

Email with AgentMail

Your agent gets its own inbox through AgentMail, an email service built for AI agents. It sorts incoming mail, groups replies into threads, reads attachments and prepares drafts. Sensitive replies wait in your drafts folder until you approve them. Address allow and block lists keep unknown senders out.

Voice mode

Speak instead of typing. Dictate a task while you drive. Ask for today’s priorities. Or think out loud about a decision and let the agent challenge you. Voice (speech-to-text and text-to-speech) is available as an add-on skill.

Scheduling and reminders

Meetings, follow-ups, renewals, tax and compliance deadlines. Tell the agent once and it reminds the right person at the right time. Recurring jobs run on their own schedule, and a morning briefing arrives before your first coffee.

Market research

Competitor moves, price changes, regulations, news in your sector. The agent runs the search, then sends a short brief with every claim linked to its source and dated. You get facts you can check, not a wall of text.

Bill management

Send a photo or PDF of a bill. The agent reads the supplier, amount and due date, logs it in your sheet, reminds you before it is due and flags duplicates or unusual amounts. It prepares. You approve every payment.

Contract drafting with your letterhead

Give the agent your template and the deal details. It drafts the contract with your company header, footer and style, and exports it to Word and PDF. It keeps the versions. A lawyer should still review anything you sign.

And more

  • Documents: create and edit PDF, Word, Excel and PowerPoint files
  • Reports and alerts on your key numbers
  • CRM and ERP connections (Salesforce, HubSpot, SAP) as an add-on
  • Omnichannel: WhatsApp, Telegram, email and web chat on the Full AI Automation plan

A day with your agent

An illustration, not a client story.

  • 7:30

    A message on WhatsApp: today’s three meetings, two overdue tasks, one bill due Thursday.

  • 9:10

    You send a photo of a supplier invoice from the car. It is logged and a reminder is set.

  • 11:00

    A client email arrives. The agent drafts the reply with the right attachment. You read, adjust one line and approve.

  • 14:00

    You say: “Draft the service contract for the new customer.” Minutes later it is in your inbox, on your letterhead.

  • 18:00

    The agent sends the day’s summary and the plan for tomorrow.

You go home on time.

Three open-source engines, one goal

We do not sell you a brand. We choose the engine that fits your case, and we can combine them. All three are open source under the MIT licence, so there is no licence fee and no lock-in to a vendor&rsquo;s code.

OpenCode OpenClaw Hermes
What it is An open-source AI coding agent A self-hosted agent that lives in your chat apps A self-hosted agent from Nous Research that learns from repeated work
Strength Builds, fixes and extends software. Read-only Plan mode, then Build mode on your approval. Parallel sessions. Messaging-first: WhatsApp, Telegram, Slack, Teams and more. Skills, memory in plain files, built-in scheduling. Persistent memory and self-written skills. Reaches 20+ platforms including WhatsApp, Telegram and email. Scheduling, subagents, approval layer.
Models 75+ providers, including local models Anthropic, OpenAI, Google, DeepSeek and local models Nous Portal, OpenRouter, OpenAI, Anthropic and local models
We use it for Your custom tools, dashboards, integrations and automations The always-on assistant your team talks to Assistants that get better at your recurring workflows
Community 160,000+ GitHub stars About 390,000 GitHub stars About 250,000 GitHub stars

Sources: ToolDirectory (OpenCode, July 2026), MindStudio (OpenClaw, 2026), Parallel (Hermes and star counts, October 2026). Community figures change daily.

OpenClaw&rsquo;s creator joined OpenAI in February 2026, and the project moved under an independent foundation. We follow each project&rsquo;s releases so you do not have to.

Right model, right price

Most tasks do not need the most expensive model. We route routine work (sorting, extracting, drafting, reminders) to fast, low-cost models and keep the powerful ones for the moments that need them.

Your plan includes a monthly token allowance: 10 million on Solo Enterprise, 25 million on Small Team, 100 million on Full AI Automation. Extra usage is billed at a published top-up rate. We explain the reasoning in our articles.

A good agent does the routing for you: it sends the simple work to the cheap model and keeps the expensive one for the moments that need it. That is one of the first things we design into every agent we build.

Security is the real work

These engines are powerful and young. Some early setups were careless, and the record shows it. In early 2026, researchers found nearly 900 malicious or dangerously flawed community skills for OpenClaw, and a critical one-click remote-execution flaw (CVE-2026-25253) was disclosed. In a 2026 survey of 750 enterprise leaders, 88.4% reported at least one AI-agent security incident in the past year.

Sources: Immersive Labs, February 2026 · AvePoint, via AI2Work.

So most of our job is not installing an agent. It is making it safe to use. On every project we put in place:

Dedicated, isolated environment

A dedicated, isolated cloud environment for your agents, never shared with another client.

No public exposure

No public exposure of the agent’s control panel, and every update applied.

Reviewed skills only

We do not install community add-ons we have not checked.

Least privilege

Each agent gets only the accounts and files its job requires.

Human approval

Human approval for payments, contracts sent outside the company and sensitive emails.

Allow lists

Allow lists for who can talk to your agent and who can email it.

Monitoring and maintenance

Monitoring and maintenance after launch, covered by your yearly renewal.

GDPR-compliant architecture

Your data is used only to deliver your project and is never sold.

Agents can still make mistakes. We design for that: they cite sources, show their work and ask before acting on anything that matters.

How it works

1

Consultation

A short, no-pressure call. You tell us what slows your business down. We tell you honestly whether an agent is the right answer.

2

Quote

Choose a plan and add only the features you need. You see the exact price before you commit.

3

Build and deploy

We configure the agent, connect your channels and tools, and test it with your real tasks.

4

Run and improve

We monitor, maintain and adjust. You can add skills, channels and agents at any time.

Plans and prices

Transparent, modular, in INR, EUR or USD. Prices exclude taxes and are confirmed after a short consultation.

Plan Best for What is included Price
Basic Automation A first automation Custom workflows, AI reporting, forms, alerts, scheduler, documentation, 30 days of support €590 one-time
Solo Enterprise One person or a micro-business 1 agent, 24/7, WhatsApp or Telegram, document skills, 10M tokens per month €1,050 year 1
Small Team Most popular A team of two or three 3 agents, 2 seats, shared workspace, permissions, 25M tokens per month €1,780 year 1
Full AI Automation A growing company Up to 8 agents, 10 seats, 3 workspaces, omnichannel, SLA, 100M tokens per month €6,220 year 1

From year 2, agent plans renew at 15% of the initial price. This covers maintenance, monitoring, updates and support. Add-ons such as email processing, voice, extra agents and CRM or ERP integration can be added or removed later.

Build your quote

Frequently asked questions

Direct answers to the questions businesses ask most often about AI agents.

No. You talk to your agent in WhatsApp or Telegram, in your own words. We handle the setup, the security and the maintenance.

Not by default. Payments, contracts leaving the company and sensitive emails wait for your approval. You can loosen this for low-risk tasks if you choose.

It depends on the task. We use low-cost models for routine work and stronger models for hard problems. Your token allowance covers normal use.

Your agents run in a dedicated, secured environment. We follow GDPR, use your data only to deliver the project and never sell it. Read our Privacy Policy.

You pay no licence fee, you are not tied to one vendor&rsquo;s code, and the engines improve every week thanks to large communities. The trade-off is that they need careful setup and upkeep. That is the part we do.

Then learn with us. Spengtra teaches practical AI classes, remote or in person in Pondicherry, Singapore and Mauritius. Same goal, two doors.