One memory layer for the tools you already use
Claude Code
Cursor
Windsurf
Codex
Cline
Kilo
Augment
Antigravity
Your AI credits run out mid-task. You jump to another tool — and start from zero. DevMemory saves your goals, decisions and code as you work, then restores them in the next tool. Deterministically.
Free tier included · Self-hostable · No vendor lock-in
One memory layer for the tools you already use
Claude Code
Cursor
Windsurf
Codex
Cline
Kilo
Augment
Antigravity
Saving your work can't depend on the model remembering to do it. So it doesn't.
Saves each turn through OS-level hooks and a watch daemon — with zero reliance on the AI calling a tool. Your work is captured before the credits die.
One devmemory continue in a new tool restores your goals, decisions and next steps — a resume prompt tuned for that tool. No copy-paste.
Not a transcript dump. Typed blocks — goals, decisions, code, errors, next steps — so the next session reads intent, not noise.
Auto-save is scoped to the one project you attach with devmemory start. Nothing saved until you say so — you're always in control.
Run it locally on SQLite or scale on PostgreSQL. Your context stays on your infrastructure. No lock-in, MIT-licensed.
We never fake support. devmemory watch --list shows the exact save mechanism per tool — hook, daemon, or MCP+rules.
One command wires DevMemory into your tool via MCP + hooks. Python or Node — your choice.
Run devmemory start in your project. It attaches, restores context, and auto-saves as you work.
Hit a wall in one tool? devmemory continue in the next — your context follows you.
Set up DevMemory in under a minute and keep your context — across every switch, every credit reset.