A local two-agent coding setup uses shared memory to preserve architecture decisions
PrajwalTomar_ · x · 2026-07-24
- The author built a fully local multi-agent setup with two agents sharing one memory: a Coder that writes code and logs architectural decisions, and a Reviewer that checks new code against that shared decision history.
- The key idea is that most agents restart from a blank slate, so the real missing piece is persistent shared memory that survives restarts and lets agents coordinate without talking directly.
- In the demo, the shared log stores decisions like SQLite over Postgres, snakecase API fields, and no external HTTP calls in the core module.
- The memory layer is implemented with Actian VectorAI DB, run locally in Docker, and used as a semantic decision log so the reviewer can find relevant rules even when the code doesn’t mention them verbatim.
More from coding & agent
- Developer turns a vintage phone into a voice agent with AssemblyAI's API — AssemblyAI · 2026-07-24
- ComfyUI package adds model-only LoRA stacking and trigger-prompt merging — boulettoxx · 2026-07-24
- Why Claude Code still uses grep: a field guide to multi-vector search — antoine_chaffin · 2026-07-24
- Anakin pitches a self-hosted web scraping layer for agent loops — Roger_M_Taylor · 2026-07-24
- Netlify’s London event promises live AI builds, agent demos and 10,000 credits — thisiskp_ · 2026-07-24
- Reddit post maps the “Periodic Table of Agent Infrastructure” — ozzyboy · 2026-07-24