Skip to main content
Back to trending

nautilus-compass

Reliability layer for multi-agent setups — keep agents coordinating without an orchestrator. Cross-dialog contracts + drift detection + black-box memory (no LLM extraction).

  • a2a
  • agent-memory
  • agentic-ai
  • claude-code
  • cross-agent
  • llm
  • local-first
  • long-term-memory
  • mcp
  • memory-layer
  • nautilus
  • rag
View on GitHub
Stars
1,255
Forks
36
+ today
—
Created
5mo

Ranking data as of October 7, 2026 (UTC).

Overview

Reliability layer for multi-agent setups — keep agents coordinating without an orchestrator. Cross-dialog contracts + drift detection + black-box memory (no LLM extraction).

The project is written in Python and has 36 forks. It was created 5mo ago.

AI-generated summary of the repository README — may contain errors.

What it is

`chunxiaoxx/nautilus-compass` describes itself as a reliability and memory layer for multi-agent setups, intended to help agents coordinate without an orchestrator. The description highlights cross-dialog contracts, drift detection, and black-box memory that does not use LLM extraction.

Described capabilities

The README preview names a workflow involving `UserPromptSubmit` for time-bucketed memory recall and drift checking, `PostToolUse` for a mid-session writer, and a `Stop` hook. The available evidence does not explain the full hook lifecycle or data format. Tags connect the project with agent memory, cross-agent coordination, and Claude Code, but do not establish integrations beyond those explicitly documented in the preview.

Getting started

No complete setup steps were supplied. Consult the repository README for supported environments, hook configuration, and details on how memory is stored and shared. The description's “without an orchestrator” phrasing is a stated design goal, not a substitute for configuration documentation.

Installation
git clone https://github.com/chunxiaoxx/nautilus-compass.git
cd nautilus-compass
# see README for setup

Frequently Asked Questions

What problems does the project target?

Coordination and reliability across multiple agents, including drift detection and cross-dialog contracts.

Does it use an LLM to extract memory?

The description specifically calls its memory black-box and says it uses no LLM extraction.