Morning briefing
Every morning the agent sums up yesterday’s new leads and stalled deals, and sends it to you as a task or to a team channel.
Custom AI agents
Custom AI agents that work by your rules.
Describe the job in one sentence and your AI agent is ready, with a name and a job sheet. You decide what it learns from, what it can touch, how much it can spend and what it does on its own. Several agents can even work a problem together — the final call stays yours.
4.7 / 5 from real users14 days of full access3,000 AI credits on usNo contract — leave anytime
Agent to agent4 / 10Stop
What should they talk about?
How do we win back customers who’ve gone quiet for six months?
Maya· SalesAI agent14 customers haven’t ordered in six months. A personal email beats another newsletter.
Jordan· SupportAI agentThree of them filed complaints. Let’s answer those first.
Ethan· MarketingAI agentFor the rest, I’d send the spring launch with a booking link.
Summary
Reply to the three complaints first; send the other 11 a personal email with the booking link.
Approve (2)Nothing is saved on its own
- A working agent from one sentence — AI writes its name and job description
- A knowledge base from your own material: files, links or Google Drive docs
- You decide what it can access — enforced by the server, not just the screen
- Its own credit budget, with a daily and a monthly cap
- 2–5 agents discuss a topic and wrap up with a summary — you approve what they suggest
In numbers
- agents discuss one topic in Agent to agent
- 2–5
- messages max — then a summary wraps it up
- 10
- knowledge sources per agent: files, links, Drive docs
- 30
A generic AI doesn’t know your business.
Every chat starts with you explaining what you sell, how you talk to customers and what you must never promise. Tomorrow you start over.
And when you do hand something off, you can’t tell what it can access, how much it’s spending, or whether it will send something in your name.
One question touches sales, support and marketing at once, but the AI only has one point of view. Pulling it all together is still your job.
A teammate with a job sheet and limits
1
Describe the job
One sentence is enough, like “Follow up on stalled proposals.” AI gives it a name and a job description, and the first conversation fine-tunes it. Everything saves automatically.
2
Teach it
Add its knowledge base — files, links, Google Drive docs, up to 30 sources per agent — and write its “How to work” and “Never” rules.
3
Set the limits
Full access, read-only, or only the tools you check off. Decide when it should ask first, which connected accounts it can use, and how many credits it can spend per day or per month.
4
Give it work
In your own words: what to do and when — every morning, on a new lead, on a new booking, when an email comes in. A test run on a real example shows what it would do — without sending anything.
What it’s used for
Three situations you’ll recognize.
New lead, right away
When a new lead lands in your pipeline, the agent reviews it and emails you what to say. The email to the lead only goes out after you approve it.
Campaign brainstorm
Your sales, support and marketing agents discuss how to win back quiet customers. After at most ten messages you get a summary and approve the suggested tasks with one click.
Why this is different
Your own agents instead of a generic chatbot.
A generic AI meets you as a stranger every morning. Your own agent knows your business, your rules and its limits.
Knowing your business
You explain it again in every chat
Knowledge base and rules on its job sheet
Access
All or nothing
Full, read-only or hand-picked — enforced by the server
Cost
No idea who uses how much
Daily and monthly credit cap per agent
Recurring work
You ask for it every time
A job: what to do and when
Several viewpoints
You merge them yourself
2–5 agents compare notes and summarize
Outgoing email
Copy, paste, send
It drafts, it goes out when you click
Companies that have already handed the work over
Don’t take our word for it.
4.7 / 5
real user rating
I’ve tried a lot of project management software, but finally here’s one that’s easy to use and meets high expectations.
I’ve been using Gammatica for over a month. We moved off Notion onto this platform quickly and painlessly.
We’ve tried many platforms, but I can finally see all our international campaigns in one place.
A 10/10 solution if you want to see what your team actually does each day.
Viktor and the team did excellent work rolling Gammatica out for us. Recommended.
Finally a calendar I don’t get lost in. I recommend it to everyone.
Anna
You don’t have to learn it. You have someone to ask.
Anna is the assistant built into the system. Not a generic chatbot: she knows Gammatica specifically, across about 300 topics — from credits to permissions to billing. Ask in a sentence and she answers; where it helps she shows you with a small animation and links you straight to the right screen.
- Ask in plain English, get a plain answer
- She doesn’t just describe where it is — she takes you there
- Available at night too, with no ticket to raise
AAnnaLive
What people actually ask her:
How do I build my own AI agent, and what can it access?
Let me open it for you
What you want to know before you commit.
No. Describe the job in one sentence and AI fills in the rest: a name, a job description, a first question. You set the role, the tone and the rules with a click or a few words. There are ready-made templates too, like a morning briefing, an instant look at every new lead, or a weekly wrap-up.
No. Sending outside your company and deleting always wait for your click, however much autonomy you give it. Set it to be more careful and it asks before every write. Tools it wasn’t granted never reach it on the server side — and it tells you so if you ask for one.
Pick 2–5 agents, type a topic, and they take turns discussing it — up to 10 messages, followed by a short summary. If an agent suggests an action (a task, a contact), an approve button sits under the summary: nothing is saved on its own. “Stop” ends it anytime, and you only use credits for the part that ran.
Yes. When you tell it something lasting in a conversation — a process, a fact about your business, or a rule like “always address customers by first name” — it decides whether to learn it and shows you what it picked up on a card. One tap undoes it. It never stores passwords, keys, card or bank account numbers.
In Gammatica AI with @name, in team chat, through its jobs, and in the “Run agent” step of an automation. If you want, it also answers on your website, on Telegram and in Slack (Slack needs your own Slack app), and it can get its own email address. There’s no WhatsApp channel at the moment.
It doesn’t go beyond its permissions or its budget: once its daily or monthly cap is used up, no new run starts until the next day (or month), and it says so. It won’t carry out actions from an email. If all you need is a chatbot for your website, the AI chatbot wizard is the simpler route.
Agents run on AI credits from your workspace’s shared balance — you don’t pay per agent. You can give each agent a daily and a monthly cap; leave it empty and there’s no cap. For Agent to agent, the estimated credit cost is shown on the start button.
Build your first custom agent.
One sentence about the job, a knowledge base from your own material, and limits you set. Anything that goes out goes out on your click.
P.S. The trial costs you nothing and asks for no card details. If it isn’t for you, you close the tab and we never charge you.