Vexum — FINOS project contribution now open (live demo inside)

[Dmtiry Grankin]

Hey team,

We propose Vexum:
Contribution issue: https://github.com/finos/community/issues/420

Core idea: capture meetings (Google Meet, Zoom, MS Teams) → build Knowledge-as-Code with sandboxed coding agents → deliver intelligence in real time.

Docs: docs.core.vexa.ai
Repo: https://github.com/Vexa-ai/vexa-core
Live Demo: core.vexa.ai

Two distinct domains — Meetings API (bots, transcription, storage) and Agents API (sandboxed CLI agents with git workspaces) — can be used fully independently, but shine together.

The meeting bots are containers, and the agents (à la Claude Code) are sandboxed in containers too — so both run really well in one runtime. CLI coding agents become scalable and safe once sandboxed.

Knowledge-as-Code is how you make coding tools as powerful on general text as they are on code. Software development was revolutionized this way — knowledge work is next. Wiki-connected Markdown is the way to encode the knowledge.

Already in the wild: Vexa runs real-time transcription at the central bank of Austria (OeNB) and Infobip, and is the real-time transcription backbone of the ASWF DNA project (participants include Sony, Disney, ILM).

Mission: Compound an organization's intelligence by implementing the knowledge-as-code paradigm on safe, scalable infrastructure with first-class access to meetings, while keeping the human in control of the action.

Live demo flow:

• Sign in with Google or Microsoft.
• Onboarding — your workspace is seeded from a knowledge graph of public FINOS data, collected autonomously by that same Vexum agent running on a schedule (https://github.com/Vexa-ai/finos-knowledge-graph). An onboarding agent asks about you so the workspace personalizes around that FINOS base.
• Start a meeting — create a Google Meet, drop the link in, and a bot joins the call. Talk to it, share a video — it's listening.
• Chat live — during and after the meeting, the agent updates your workspace, so it can hand intelligence back to you in real time.
(Backend, for the curious: a Claude Code agent spawns in a container to manage your workspace; the meeting bot is a container too — same runtime, both sandboxed.)

Best,
Dmitry Grankin
Vexa.ai