Show HN: Jarvis-AI, an AI Agents network that kills admin work in big corporate

6 oliomap 2 4/30/2025, 6:49:47 PM github.com ↗
Cheers HN,

We're Oli and Pascal, two friends from ETH Zürich. We built a network of AI Agents for large organisations, that finally gets rid of all admin work for employees. Current features are:

- Schedule, move or cancel meetings (via Google Calendar or locally)

- Dynamically adapt meetings according to stakeholders’ availabilities (internal communication of the agents)

- Summarize incoming mails (via Gmail)

- Create a project plan (command: plan XXX = [project description]) including stakeholders, timeline and cost estimate

- Plan, assign and view tasks

- Do all of the above via audio

We believe that the network point is the core of the product. If you're planning a new project, Jarvis should not only give ideas but also propose whom to work with based on the context information of all the other Jarvises in the company. Instead of sending mails, information flow happens between the Agents and the audio feature makes it super natural to speak to your Jarvis.

This is very early stage, so any advice/feedback is much appreciated :)

Comments (2)

badmonster · 5h ago
how is scheduling conflict resolution handled under the hood — is it hard-coded logic or learned via prompting?
oliomap · 9m ago
Currently it resolves the conflict via an LLM-call. The plan is to replace that call with a communication routine between the responsible nodes, e.g. also a prompt but instead of directly to OpenAI, to the other node. They should then communicate until resolving the issue and finding a better solution. As soon as we built a personalization and a memory, this should be doable, right?