Skip to main content

​
Claude Code Setup

Claude Code forgets everything between sessions. Atlas fixes that.

​
What you get

  • Claude remembers your stack, team, and architecture decisions
  • No more re-explaining context at the start of every session
  • Decisions made with Claude get automatically saved for next time

​
Step 1 — Add to CLAUDE.md

Create or update CLAUDE.md in your project root: 
## Persistent Memory (Atlas)

You have access to Atlas cognitive memory for this project. 

**Always search memory before answering questions about:**
- Architecture decisions
- Technology choices  
- Team assignments
- Past bugs and their solutions
- Deployment configuration

**Save to memory when:**
- A new architectural decision is made
- A tech choice is finalized
- A team assignment changes
- A significant bug is resolved

Use these commands to interact with memory:

\`\`\`bash
# Search
python scripts/memory.py search "your query"

# Save
python scripts/memory.py save "your fact"

# Graph QA
python scripts/memory.py ask "your relational question"
\`\`\`

​
Step 2 — Seed your project memory

Run this once to give Claude Code context about your project: 
from atlas_memory import CognitiveBrain

brain = CognitiveBrain(
    api_key="atlas_your_key_here",
    user_id="my-project",
)

# Seed project knowledge
facts = [
    "The frontend is built with Next.js 16 and deployed on Vercel.",
    "The backend API uses FastAPI and is deployed on AWS ECS.",
    "The primary database is PostgreSQL hosted on AWS RDS.",
    "Alice is the backend lead. Bob is the frontend lead.",
    "Authentication uses JWT tokens stored in HTTP-only cookies.",
    "The CI/CD pipeline runs on GitHub Actions.",
]

for fact in facts:
    brain.add(fact)

print("Project memory seeded.")

​
Step 3 — Session workflow

Start every Claude Code session with: 
# What has changed recently?
python scripts/memory.py search "recent changes last week"

# Check current architecture
python scripts/memory.py ask "What is the overall system architecture?"
End sessions by saving what was decided: 
python scripts/memory.py save "Decided to use Redis for session caching instead of DynamoDB on 2025-04-12."

​
Example prompts that work well

Once memory is seeded, Claude Code can answer:
“What database are we using and why?"
"Who should I talk to about the authentication service?"
"What deployment platform are we on?"
"What have we decided about caching?”