Topologies

A topology is the shape of a Naylence fabric: where nodes run, which nodes host agents, and how messages flow between them.

You can build the same logical system in many shapes. Topologies let you start small (single process) and grow into multi-node, multi-runtime deployments without changing the core mental model.

What changes when you change topology

Topologies are mostly about placement and boundaries:

Where do nodes run? (same process, separate processes, browser vs backend)
Where do agents live? (on a sentinel, on dedicated agent nodes, on client nodes)
Where is policy enforced? (admission, routing rules, trust boundaries)
What are the failure/latency boundaries? (local calls vs network hops)

The primitives remain the same — fabric, nodes, agents, clients, and sometimes sentinels. If you haven’t read those yet, start here:

Concepts → Overview

A simple mental model

Think in layers:

Client nodes initiate work and can also listen/stream.
Agent nodes host addressable capabilities.
Sentinels (when used) sit at boundaries and concentrate admission + routing policy.


[ client node ]  ◄──►  (sentinel)  ◄──►  [ agent node(s) ]

Not every topology needs all three.

Start here

If you are new to Naylence, pick the smallest topology that matches your goal.

Single-process fabric

One process, one node. Best for learning the API and debugging without networking.

Agent on sentinel

A sentinel that also hosts your first agent(s). You get admission + routing early with minimal moving parts.

Dedicated sentinel + agent nodes

Separate control (sentinel) from work (agents). Best when you want isolation, scaling, or mixed runtimes.

You can move between these topologies incrementally. The point is not to pick the “best” one up front — it’s to pick the smallest one that fits your current constraints.

Live demos referenced in this section

These pages link to small, intentionally-boring demos where the shape is the main thing to learn:

Hello (agent on sentinel): examples-ts-hello
Hello (dedicated sentinel, 3-node): examples-ts-hello-3node
Multi-agent workflow (browser-only): examples-ts-multi-agent

How to choose (quick decision guide)

You want the simplest possible start → Single-process fabric
You want a “real” front door early (routing/admission), but still minimal → Agent on sentinel
You want isolation, scaling, or mixed runtimes → Dedicated sentinel + agent nodes
You want a zero-infra demo or rapid orchestration prototyping → Browser-only fabric

Next steps

Pick a topology above and follow the page.
Once you’ve run one, continue with Guides for step-by-step buildouts and operational topics.