Project Memory
The foundational use case: persistent, project-scoped memory for AI coding agents. Your agent remembers preferences, decisions, architecture, and context across sessions — and consolidates them into generalized understanding.
The problem
Every coding session, your agent starts fresh. You tell it your conventions, explain the architecture, re-state your preferences. It's like onboarding a new developer every morning.
Setup
Install Remind and the base skill in your project:
pip install remind-mcp
cd your-project
remind skill-installThis creates .claude/skills/remind/SKILL.md and sets up the project-local database at .remind/remind.db.
Walkthrough
Session 1: Initial context
# Agent stores what it learns
remind remember "Project uses React 19 with TypeScript, Vite for bundling"
remind remember "API is FastAPI with SQLAlchemy ORM" -t observation \
-e tool:fastapi -e tool:sqlalchemy
remind remember "Chose PostgreSQL over MySQL: better JSON support, team familiarity" \
-t decision -e tool:postgres
remind remember "All API endpoints must validate input with Pydantic" \
-t spec -e tool:pydantic -e module:api
remind remember "User prefers explicit error handling over exceptions" \
-t preference
# End of session
remind end-sessionAfter consolidation
Remind consolidates these into concepts like:
"Project tech stack centers on Python (FastAPI + SQLAlchemy) backend with React/TypeScript frontend"
- Confidence: 0.9
- Relations: implies → "Team has Python backend expertise"
- Conditions: "Current project architecture"
"User values explicitness and type safety across the stack"
- Confidence: 0.8
- Source: preference for explicit error handling + TypeScript usage + Pydantic validation
- Relations: correlates → "Prefers statically typed languages"
Session 2: Context carries forward
# Agent recalls at session start
remind recall "project architecture and tech stack"
remind recall "user preferences and conventions"
remind decisionsThe agent immediately knows the stack, the conventions, and why decisions were made — without re-explaining anything.
Session 5: Knowledge deepens
After several sessions, consolidation has produced a rich concept graph. New episodes get integrated with existing knowledge:
remind remember "Added Redis caching layer for expensive DB queries" \
-t decision -e tool:redis -e module:cachingThis consolidates into the existing tech stack knowledge, possibly creating:
"Project follows a pattern of adding infrastructure tools (Postgres, Redis) based on specific performance needs rather than speculative scaling"
- Relations: specializes → "Team pragmatic about infrastructure choices"
What you get
After a few weeks of use:
- No more re-onboarding — The agent knows the project cold
- Decision audit trail — Every architectural choice is remembered with rationale
- Growing understanding — Not just facts, but patterns and preferences
- Entity graph — Navigate from a file to its specs, from a tool to the decisions that chose it