Skip to content

Coming soon

Everything you organize, folded into one careful agent.

For small businesses and for people running their own lives and projects. In origami, an AI agent works on your mail, files and services. It reaches them only through origami, which holds every key, checks every call and writes everything down.

See how it worksIn development. The first release is coming soon.
A white paper origami swan
  • Keys never reach the agent
  • Every change previewed
  • Every call in the audit log
  • Your notes stay plain files
origami on its Dashboard: a sidebar with Agent, Organization and Broker sections, today's brief with four tasks tagged by project, and a list of four projects with their goals and due dates.
origami's Dashboard: today's brief, the tasks across projects and the projects at a glance. Sample data.

Who it is for

For your business and for your own life.

origami is made for people who run things themselves, not for large companies. A studio of six and a household of two get the same careful agent.

Small businesses

Mail, documents, invoices and what the team knows, in one place. The agent does the routine and you approve what goes out.

  • answer customer mail
  • file receipts and invoices
  • keep procedures for the next person

Private life and your own projects

The letters, bills and deadlines of a household, and the projects you run besides it: a move, a renovation, a trip, a club. Same care, same keys kept safe.

  • sort the letterbox and the bills
  • plan a move or a renovation
  • keep track of your ideas

How it works

One way in. Every step on the record.

  1. In origamiYou and the agentYou ask, it works. It holds no keys.
  2. origami broker
    • VaultEncrypted, opened for the working day.
    • PolicyDecides: runs, asks, grant or refused.
    • AuditLogs every call. Daily caps on sending.
  3. Your systemsMail, calendar, files, lettersSee only the broker.

Keys stay in the vault

Credentials live in an encrypted vault that the broker opens for the working day. They never reach the agent, your files, the logs or a prompt.

You approve what you saw

A change shows its exact preview first. If anything in it changes before the call runs, your approval is void and the agent has to ask again.

Nothing happens off the record

Every call, verdict and approval lands in the audit log, with daily caps on things like sending e-mail.

Policy

Four verdicts. No grey areas.

Every call the agent makes gets exactly one, decided by the broker before anything runs.

  • runsReads go straight through.search mail · list files · read a document
  • asksChanges wait for you, with an exact preview.send an e-mail · create an event · send a letter
  • needs a grantRisky work opens for a limited time, with a reason. Each call still asks.run a command on a server
  • refusedSecrets, access settings and bypasses never run.read a secret · change access rights · use gh or gcloud around origami

Built in

What the agent knows stays in plain Markdown.

Knowledge, procedures, projects and ideas are plain Markdown files that belong to you. Read and edit them in origami or in any editor, and keep them in git if you like.

my-origami/
├── knowledge/    # what you know
├── sops/         # how things are done
├── projects/     # goals and tasks
├── ideas/        # what might come next
└── origami.yaml  # tools and settings

Let the agent do real work. Keep the keys.

origami is in development at syncrea GmbH. The first release is coming soon.