Canonical project state
Memory, requirements, plans, soul lessons, AI docs, tests, and relations live in one system.
AI infrastructure
MemoryAndSoul
A persistent database for AI-agent work: memory, requirements, plans, role lessons, architecture docs, tests, and context retrieval in one system.
Problem
AI coding sessions lose context when a window fills up, a worker returns, or a project spans many days. Notes in flat files drift, requirements blur, and lessons get rediscovered instead of reused.
Solution
MemoryAndSoul makes project memory canonical in a database and exposes it through MCP tools. Agents can list current work, persist durable facts, save role-specific lessons, track requirements, and retrieve focused context.
Context packs
Agents can start work with bounded, relevant context instead of rereading the world.
Audit-friendly flow
Lifecycle transitions and durable notes preserve what happened and why.
Technical highlights
Architecture
- PostgreSQL-backed entities for memory, soul, requirements, plans, docs, tests, and relations.
- MCP tool surface for AI agents and local automation.
- Immutable/append-oriented memory design to preserve history.
- Project-scoped and global rows for reusable cross-project learning.
Dave's role
Founder, product designer, and systems architect defining the persistence model, agent workflows, safety rules, and cross-session project discipline.
Stack and status
Stack
- Python service layer
- PostgreSQL
- MCP tools and local client integrations
- Structured requirements, docs, and agent workflows
Private repo / active system
Open GitHub
Repository access is available to Dave when logged in.