PRAW
Book a call

AI agents, tailored to your business

Own your AI agent instead of renting one.

We set your agent up on an account in your name and teach it how your business works. If you ever stop working with us, it keeps running. This page covers the details, including the technical ones.

Book a call See how it works

your agent

Agents

  • Order desk@order-deskliveAnswers order and shipping questions in WhatsApp.
  • Dispatcher@dispatchliveBooks jobs and reshuffles the crew's week.
  • Back office@back-officeliveMatches payouts and chases unpaid invoices.

Agents: Each agent has one job, described in plain words

Example data

It grew out of real work.

Proven on a live business first

The first agent we set up runs a retail consultancy's client work every day. What held up in real use became the way we build every agent since.

Real workWhat held upYour agent

Built on an open foundation

The software underneath is open source: free to use and public, so anyone you trust can check exactly what it does. We did not build a black box and ask you to trust it.

Ownership, not just privacy

Plenty of tools promise your data stays private while the rules, the software and the decisions about what comes next stay with them. Here every layer is yours.

PlaybookDataModelsCodeServer

Put the judgement of your best people into the system.

Every business runs on the judgement of a few experienced people, and growing means finding more of them. An agent you own carries that judgement for you.

  1. Procedures become skillsHow a job is done, step by step, in your words.
  2. Standing facts become memoryLimits, owners and policies, kept current.
  3. Every agent works from bothChange a rule once and every agent follows it.
Store owner, handover note
“If a customer wants a refund, check the order and the tracking first. Anything over $200 comes to me.”

Written down once

Skill

skills/refund-request.md Handling a refund request
  1. Find the order
  2. Check the tracking
  3. Route by amount

Memory

memory/refund-policy.md
Refunds up to $200
Agent
Above $200
Owner approves
Policy version
v3, March

Used byOrder deskReturns helperWeekly report

Example data

Your agent

noun: an AI worker your business owns

A worker you own, not a chatbot you rent.

An open-source AI agent that runs on your own server, works from your playbook and your data, and does real jobs across the business. The setup, the code and everything it learns belong to you.

Most business AI is rented.

Your prompts and data may stay yours, while the code, the rules and the roadmap stay with the vendor. You furnish the house and someone else owns it.

With your own agent you hold every layer, including the one that does the work.

Chatbot subscription

  • PlaybookRented
  • DataRented
  • ModelsRented
  • CodeRented
  • ServerRented

0/5 layers yours

Automation platform

  • PlaybookYours
  • DataYours
  • ModelsRented
  • CodeRented
  • ServerRented

2/5 layers yours

Your agent

  • PlaybookYours
  • DataYours
  • ModelsYours
  • CodeYours
  • ServerYours

5/5 layers yours

Tap or point at a row to see what it means.

Your agent belongs to you.

The setup, the code, the data, the server and the models. Nothing needs our permission to keep running.

Any model, changed in one place

Use a model from any major provider, or one running on your own hardware. Change it once and every job the agent does follows.

  • Claude
  • GPT
  • Gemini
  • A local model, on your hardware

Your data stays on your server

Conversations, skills, memory and every record of what the agent did live on your machine.

ConversationsSkillsMemorySchedulesKeys

Open source. Read every line.

The code is public under the MIT licence. Anyone you trust can read all of it before we start, and any developer can change it after we leave.

Runs where you choose

It runs on the machine you choose. The only information that leaves it is what goes to the AI model and the tools you pick.

Your cloud accountA box in your officeA small rented server

The tools you already pay for

It connects to your store, inbox, calendar and books, so there is nothing to migrate.

One agent, wired into everything you run.

It reads and acts in

Your agent How you do things What it has learned

Tap a box to see what the agent does with it.

yours to keep

So it can answer

From install to agents at work.

$ install agent

gatewayrunning

memoryrunning

schedulerrunning

Ready on your server

We install it on your account

We put the agent on your server or in your cloud account, lock it down and hand you the keys.

Instructions

Model

any provider

Tools

StoreEmailSend refund needs approval

Teach it in plain language

You describe the job in plain words: what to do, which tools it may use, and what needs your sign-off first.

One agent

TeamsWhatsAppTelegramSlackDiscordSignal

Run it where people work

The same agent answers in the chat apps your team and your customers already use, with the same rules everywhere.

As your agent takes on more, one set of rules keeps it in line.

One place to teach it and one record of what it did: what ran, what it touched and who allowed it.

Know-how, written once

A procedure becomes a skill in plain language. Edit it, and the next answer uses the new version.

Skill

skills/month-end-close.md # Month-end close checklist

Used byBack officeWeekly report

Work that runs on a clock

A job that runs at 7:00 every morning follows the same limits, and leaves the same record, as one you asked for in chat.

  • Daily numbers7:00To your Teams chat.
  • Weekly reviewFri 16:00What it could not answer.

A person signs off before it acts

Anything that spends money or reaches a customer can wait for your yes, given right in the chat.

Waiting for you

Send quote to Dalton, $6,450

ApproveDecline

See it on your own business.

A 30-minute call. We take one of your jobs, map it to an agent, and tell you plainly whether it is worth doing.

Book a call