AI: agent-operable software
Vivotics
Project management you can run from a chat window
The problem
Delivery teams do not live in their project tool. They live in chat, in calls and in the work itself. So the tool is always out of date, and the status report is written by the person who least wants to write it.
What we built
- A multi-tenant platform: projects, tasks and clients with per-company isolation, role and module gating, and a client portal with its own permission surface and per-client notification control.
- An MCP server that exposes the workspace to any MCP client. Connect Claude Code, Claude Desktop or a generic stdio client with a scoped token and ask it to plan, update or report. Setup is a token, one command and a config file, with documented revoke and troubleshooting paths.
- An automation engine with real triggers. Events in the workspace fire workflows, so a finished meeting or a changed task state moves work forward without anyone opening the app.
- An AI notetaker. A bot joins the call, the audio is transcribed, a model drafts the minutes, and action items are extracted as suggested tasks linked back to the meeting. The transcript stays searchable segment by segment.
- An AI writing assistant inside the workspace for briefs, replies and documentation, plus meeting scheduling with automated reminders.
- Queue workers and a scheduler running as managed services, so background work is observable rather than buried in a cron file.
Where the automation sits, and where it does not
The model drafts, extracts and summarises, and every meeting output lands as a suggestion that a human accepts before minutes become tasks. The model never sequences the work: a confident wrong answer about what to do next costs a client relationship, so prioritisation stays with a person. The system makes the facts easy to see; it does not decide.
What we would not claim
Vivotics is our own platform. It is not sold as a product yet and has no external customer base we can cite.
Live in production since July 2026.
Your turn
We set the baseline before the work starts.
That is the part that makes a number mean something a year later.