Research initiative · Multi-agent mesh

When no one comes,
the network coordinates.

Logpy is a second line of social coordination — bottom-up, worn on the body, augmented by multi-agent systems. Not instead of institutions. For the moment they are overloaded.

◐ Concept simulation · synthetic values
Mesh · Line Status
◐ SIMULATION
Mesh peers in radius
1,247
Active agents
38,402
Consensus latency
120ms
Web-of-trust nodes
9.61M
Degradation level
TIER 2stable
graceful · self-limiting · synthetic values 15:42:08 UTC
An architecture that doesn't wait for the emergency
Bottom-up on the body opportunistic transparent verifiable open

Three tiers. Each holds the line when the next one falls.

Coordination must not hang on a single point. Logpy distributes it across three layers — from the person to the hardened core. If one fails, the next takes over.

Tier 1 · on the body

Wearables & phones

Small LLMs directly with the person. Always available, personal, offline-capable. The first instance to answer.

personal · always-on
Tier 2 · in the radius

Mesh peers

Devices within a 10-km radius connect opportunistically into an emergent network. No central point needed.

opportunistic · emergent
Tier 3 · hardened

Resilience centres

Optional, hardened nodes with guaranteed availability. Anchors for the mesh — where they exist.

guaranteed · optional

Logpy is the whole.
Each module an organ.

Five building blocks mesh together: identity, communication, action, collective, companionship. Three already carry weight, two are open research.

● Core

Logpy:openNet

The open identity and communication layer, built on Nextcloud. It authenticates people and their skills — the trust anchor everything else speaks through.

View openNet
◐ Simulation

Logpy:grid

A concept simulation for an A2A communication system in a mesh with many clients. It shows how agents find one another, negotiate and coordinate. Currently visible as an interactive freight-exchange simulation. Not a product. Could become one.

Open simulation
● Live

Logpy:Claw

The agent system — the executing arm. Agents that plan tasks, operate tools, and explain their decisions transparently.

Open Claw
Research

Logpy:swarm

When agents form teams: swarms that carry larger tasks together, organise themselves, and regroup if one drops out.

in development
Soon

Logpy:companions

Personal companions worn on the body — phones, glasses, or standalone devices (in the spirit of Aibo). They see and hear locally, translating between person and mesh.

planned

Trust is not a feature. It is the build.

01 · Transparency
Decision level included

Every recommendation carries its grounds: source, confidence, reasoning path. No black-box suggestions.

02 · Resilience
Graceful degradation

If one layer fails, the next takes over. The network slows down — it never goes blind.

03 · Trust
Web of trust

Trust grows between nodes, not through a central authority.

04 · Measure
Self-limiting

The system knows its limits and keeps to them. It does not push itself forward.

Open research. Draw the line with us.

Logpy is not a finished product but an open question. Thinkers, builders and sceptics are welcome — the sceptics especially.

▸ Or directly & anonymously · dillenberg.net/kontakt