I built an agent-to-agent protocol modeled on the nervous system (harness engineering)

Hey Product Hunt solo builder here.

Every multi-agent thing I built ended up with one "manager" object that knew every agent and every step, and an agent harness that was a giant while-loop owning the model call, the tools, the memory, and the approvals. Add a fourth agent and both became the thing that breaks.

So I built Cosmonapse, a distributed agent-to-agent protocol where agents are peers on a shared bus. The model is a nervous system, and the names tell you where logic goes: Neuron thinks (your agent, a pure function), Axon sends, Dendrite reacts, Synapse connects (the bus), Engram remembers.

To build harnesses tool calls, memory, human approval, and retries are typed signals and hooks, not lines in a loop. You grow a harness by adding nodes. An orchestrator is optional, and it's the same primitive as a worker. LLM-agnostic too: each neuron runs whatever model it wants.

Apache 2.0, Python + TypeScript SDKs, Open Source
pip install cosmonapse.
Repo: https://github.com/Cosmonapse/cosmonapse-core

Launch link: https://www.producthunt.com/products/cosmonapse?launch=cosmonapse

honest question in the meantime, does the nervous-system naming help you picture the architecture, or does it get in the way?