Open-Source Cousins Framework Gives Claude Agents Persistent Memory and Personalities
bomba_____ · reddit · 2026-10-07
Inspired by the Sicilian "call my cousin" problem-solving culture, a developer living in Denmark open-sourced cousins-framework: a family of Claude agents (Agent SDK) each with a name, a job and its own memory — a daily generalist, one who knows PCBs and datasheets, one who cooks — that refer questions to each other.
The hard part: surviving session end
- Each agent writes a handoff before its session ends
- Memory keeps truth levels: what the owner said vs. what a tool measured vs. the agent's own guess
- Agents wake up the next day as themselves
- A web console lets you watch them work, hold meetings and browse their memory
Python + SQLite + Docker, Apache-2.0; runs on Claude or opencode's free models (git clone + docker compose up). The author is honest that persistence isn't free — heartbeats and handoffs eat tokens.
More from coding & agent
- Glasser aggregates 1,558 data APIs from 29 providers for pay-as-you-go AI agents — goyalshaliniuk · 2026-10-07
- MIT paper: a minimalist agent loop that passes history as code variables beats Letta and ACE at half the cost — rohanpaul_ai · 2026-10-07
- A 3Blue1Brown-Style Explainer Rendered End-to-End in Rust by franken_manim — doodlestein · 2026-10-07
- Netlify AI Gateway Example: A Function That Returns Nano Banana-Generated Images — thisiskp_ · 2026-10-07
- Opus 5.5 builds 'black slopboxes' that beat human code on speed and quality despite ugly aesthetics — burny_tech · 2026-10-07
- Free official AI courses from Anthropic, Google, OpenAI and 7 more, in one list — nikola_mr64990 · 2026-10-07