The difference is whether anything changed afterwards.
Delegation · real tools · a shared memory · verified output
A chatbot answers and stops. An agent is given a job, uses tools to do it, and leaves something behind — a list that exists, a document that was written, a status that moved. The test is simple: after the conversation, is there anything you can open?
adris agents drive a real browser, read your spreadsheets, write files, and record what they found. When one of them says a contact was verified, it means a page was opened and read, not that a plausible email address was produced.
Specialists, not one generalist.
Research
Finds companies and people that match a brief, and checks they are real.
Outreach
Drafts a message per person from what was actually found about them.
Data
Filters, de-duplicates and merges the sheets you already have.
Content
Posts, briefs and long-form written to a specific audience.
Finance
Pricing and unit-economics questions, worked through with your numbers.
Security & legal
Contract review and threat monitoring, handled locally.
A chief-of-staff agent decides who a request belongs to and hands it over, so you describe the outcome rather than choosing the specialist.
What one agent learns, all of them know.
Everything the team finds is written to a shared knowledge graph — your lists, your documents, your prior decisions. The next request starts from that instead of from nothing, which is what stops you re-explaining your business at the top of every conversation. See how the Brain works →
On your machine, with your data.
The agents run inside the desktop app. Your lists, your credentials and their memory are stored on your own disk. Connected accounts are called directly from your machine to the service, with the token held in an encrypted local store.
Against a chatbot, and against hiring.
| adris agents | An AI chatbot | An agency | A junior hire | |
|---|---|---|---|---|
| Leaves something behind | A file, a list, a draft | An answer in a window | A deliverable | A deliverable |
| Verifies what it claims | Opens the page and reads it | No | Usually | Usually |
| Monthly cost | ₹0–4,999 | ~₹2,000 | ₹50,000+ | ₹25,000+ |
| Where your data sits | Your machine | Their cloud | Their systems | Your systems |
| Available at 2am | Yes | Yes | No | No |
| Judgement on hard calls | Needs your review | No | Yes | Growing |
Common questions
How many agents are there?
Around 48 across ten departments, led by a chief-of-staff agent that routes work to the right one.
Can they actually send emails and messages?
They draft and prepare them, and you approve before anything goes out. Connected accounts are called from your own machine.
Do they make up contacts?
The verification step opens the page and reads it. A field that could not be confirmed is left blank rather than filled with a guess.
Does this need a cloud AI model?
No. The agents can run on a local model, on your own API key, or on included credits.
Will it send messages without asking me?
No. It drafts and prepares; you approve before anything goes out.
How does it avoid inventing contacts?
The verification step opens the page and reads it. A field it cannot confirm is left blank rather than filled with a plausible guess.
Can it work from a spreadsheet I already have?
Yes. Point it at a sheet in the Brain and it reads it, filters it and works down the rows.
How many agents are there?
Around 48 across ten departments, with a chief-of-staff agent routing work to the right one so you describe the outcome rather than picking a specialist.
Does this need an expensive AI plan?
No. Agents can run on a local model, on your own API key, or on included credits.
Is it useful for a one-person business?
That is the main case. The agents do the volume work a first hire would do, before there is a first hire.
Related
Put the first job on someone’s desk.
Krew is in the free tier. Ask for something specific and watch which desk it lands on.