Ask Claude to Save Its Progress to a File Before You End a Session
Sessions end — rate limits hit, you need to reboot, or you just need to switch tasks. But context dies with the session. Before you close, ask Claude to dump its current state to a file so the next session starts warm.
Before we stop, write a progress file at .claude/progress.md summarizing:
- What we accomplished this session
- What's currently in progress (half-done changes)
- What still needs to be done
- Any decisions we made and why
- Files that were modified
Claude writes a structured handoff document that captures everything the next session needs:
# Session Progress — 2025-03-28
## Completed
- Migrated UserService to async/await
- Added input validation to all API endpoints
- Fixed the N+1 query in OrderController
## In Progress
- PaymentService refactor — started but tests aren't passing yet
- The Stripe webhook handler needs error retry logic
## Remaining
- Update API docs to match the new validation rules
- Add rate limiting to the public endpoints
## Decisions
- Chose Redis for rate limiting over database counters (faster, team agreed)
- Kept the legacy payment callback for backward compatibility until v3
## Modified Files
- src/services/UserService.ts
- src/controllers/OrderController.ts
- src/middleware/validation.ts
Start the next session by pointing Claude at the file:
Read .claude/progress.md and continue from where the last session left off.
Start with the in-progress items.
This also works for team handoffs:
I'm handing this task to a colleague. Write a handoff document at
.claude/handoff.md that explains what I've done, what's left,
and any gotchas they should know about.
Add the progress file to .gitignore if it's personal working state, or commit it if you want the team to see it.
Context is the most expensive thing to rebuild — save it to a file before you lose it and every future session starts five minutes ahead.
via Claude Code
Log in to leave a comment.
Set up Claude Code as an automated reviewer in your CI pipeline — on every pull request, it reads the diff, checks for bugs, security issues, missing tests, and convention violations, then posts its findings as a PR comment. Your human reviewers get a head start because the obvious issues are already flagged before they look.
Before deploying, tell Claude to read your project — migrations, environment variables, queue workers, scheduled tasks, caching, third-party integrations — and generate a deployment checklist that's specific to your app. Not a generic "did you run migrations?" list, but one that knows YOUR infrastructure and catches the things YOUR deploy can break.
Instead of writing a README from memory or copying a template, tell Claude to read your project and generate one that's actually accurate — real setup instructions from your config, real architecture from your directory structure, real API examples from your routes, and real prerequisites from your dependency files.