Tutorial 5 of 771% complete
All tutorials
5 minBeginner

Session Notes

Track progress and blockers between sessions

Session Notes & Short-term Memory

Track blockers, progress, and context within your AI assistant's short-term memory.

What are Session Notes?

Continuity provides both long-term synthetic memory (decisions) and short-term session memory (session notes). While long-term memory preserves architectural history, short-term memory keeps your AI assistant aligned with your immediate goals during a single coding session.

Use short-term memory to:

  • Keep the AI focused on current goals
  • Track active blockers as you hit them
  • Record attempted approaches (what worked and what didn't)
  • Prepare a "handoff" for your next session

How to Use Them

Set session goals

You: "Note that today I'm working on fixing the login bug and adding rate limiting."

AI: "Added to session goals: Fix login bug, add rate limiting."

Track blockers

You: "I'm stuck. The OAuth callback keeps returning 401. Add this to blockers."

AI: "Added blocker: OAuth callback returning 401."

Record what you tried

You: "I tried changing the redirect URI but that didn't help. Note that in the attempted approaches."

AI: "Noted: Changed redirect URI — didn't resolve 401 issue."

Save progress

You: "Found it! The redirect URI was wrong in the production config. Note that as the fix."

AI: "Noted: Fix found — redirect URI incorrect in production config."

Get a summary

You: "Summarize what we did today."

AI: "Session summary:

  • Goal: Fix login bug
  • Blocker: OAuth 401 errors
  • Tried: Changing redirect URI (didn't help)
  • Fix: Production config had wrong redirect URI"

Session Notes Sections

Ask your AI to add notes to specific sections:

SectionWhat to track
GoalsWhat you're trying to accomplish
BlockersProblems you're stuck on
ApproachesWhat you tried (worked or not)
DecisionsQuick notes (promote to real decisions if important)
Next stepsWhat to do next
QuestionsOpen items to figure out

Example phrases

  • "Add to session goals: Implement caching"
  • "Note this blocker: Tests failing intermittently"
  • "Log that we tried X but it didn't work"
  • "Add to next steps: Review PR tomorrow"

Short-term vs. Long-term Memory

Both types of memory are part of the Continuity system, but they serve different purposes:

Short-term (Session Notes)Long-term (Decisions)
Temporary, for current workPermanent, for future reference
Quick notes and progressDocumented architectural reasoning
"I'm currently stuck on X""We chose X over Y because of Z"

Promoting to Long-term Memory

If a temporary choice becomes a lasting project direction, "promote" it:

You: "The Redis approach worked well. Let's make that a permanent decision in our synthetic memory."

AI: "I'll record decision-23: Redis for caching — fast TTL support, needed for session storage."

Starting a New Day

Begin sessions by checking notes when native tools are mounted. If not, open the repo-local session notes directly:

You: "What were we working on? Check the session notes."

AI: "From your last session:

  • Goal: Rate limiting
  • Blocker: None open
  • Next steps: Implement middleware, test with Redis"

Workflow Example

Morning

You: "Check session notes and tell me where we left off."

AI: "Last session you were debugging auth issues. Blocker was OAuth 401 errors."

During work

You: "I fixed the auth issue. Remove that blocker and note the fix."

AI: "Removed blocker. Noted: Auth fixed by correcting production redirect URI."

End of day

You: "I'm stopping for today. Save where we are and what's next."

AI: "Session notes updated:

  • Completed: Auth fix
  • Next: Start rate limiting implementation
  • Question: Should we use Redis or in-memory for rate limits?"

Key Phrases

What you wantWhat to say
Add a goal"Add to goals: X"
Add a blocker"I'm stuck on X" or "Add blocker: X"
Remove a blocker"Fixed it" or "Remove blocker"
Note an approach"I tried X" or "Note that X didn't work"
Get summary"Summarize the session"
Check notes"What's in the session notes?"

Key Takeaways

  1. Ask explicitly — Tell your AI what to add to notes
  2. Use sections — Goals, blockers, approaches, next steps
  3. Check at session start — "What were we working on?"
  4. Save at session end — "Save where we are"
  5. Promote important items — Turn lasting choices into decisions
  6. Expect degraded mode — If native tools are unavailable, use the repo-local notes and handoff files

← Semantic Search | Memory Access (MCP) →