Every context, connected.
Every agent, empowered.
Contextberg is a local memory app for your AI agents. It watches your screens, agent transcripts, and browser in the background — so Claude Code, Cursor, and OpenClaw can just remember.
You don't have to type what you already did.
Contextberg records your screens, inputs, browser activity, and agent conversations — then serves it all to your coding agent via MCP. No config files. Just connect.
- Screenshots across every window
- Your browser history
- Claude Code, Cursor, Terminal transcripts ingested
- MCP server built-in — plug in any compatible agent
The way you work, turned into memory.
Contextberg quietly watches your work and writes it into three kinds of memory — activity memory of what you did, daily memory grouped by date, and long-term memory that captures the tools you reach for and the patterns in how you work. All ready for your agent to reason over.
- Long-term memory that captures your tools & work patterns
- Activity memory that logs what you actually did
- Daily memory that aggregates progress by date
Work history stays local. Models stay flexible.
Your work history stays local. Use your existing Codex sign-in, a Gemini/OpenRouter API key, Contextberg Cloud, or a local model.
The moment you're back, you're back in.
Contextberg shows you exactly what you were doing before you stepped away — auto-summarized from your recent activity, browser history, and agent transcripts. Anything you want to dig into, just ask in chat.
- An auto-summary of your last work, ready the moment you return
- Reconstructed across activity, browser, and agent history
- Dig deeper on anything, right there in chat
“Pick up where I left off.”
No more spelling out filenames or open tabs — your agent already has the context to respond.
Bring your context to
your agent.
Install in two clicks. No account. No cloud lock-in.