Contextberg

Local AI agent memory for macOS & Windows, served via MCP

145 followers

Local AI agent memory for macOS & Windows, served via MCP

145 followers

•

•

Contextberg is a local AI agent memory app for macOS and Windows. It captures screens, browser history, and agent conversations into a private, searchable archive, then serves relevant context to Codex, Claude Code, Cursor, and other agents via MCP. Search screenshots with OCR, exclude any app or source, and use your existing Codex sign-in, Gemini/OpenRouter API key, Contextberg Cloud, or a local model. Your archive stays on-device; only selected context goes to your chosen provider.

Free

Launch tags:Productivity•Time Tracking•Artificial Intelligence

Launch Team / Built With

LightfieldAI-native CRM that builds itself and does work for you

Promoted

The MCP angle is smart , instead of re-explaining context every session, agents just pull what they need. Curious how Contextberg handles sensitive data that shows up in screen recordings, like passwords or private chats. Also wondering if there's any control over what gets stored vs ignored. This could genuinely change how I use Claude Code day-to-day.

Report

3mo ago

Contextberg

Maker

@tejasvi_singh_jodha User control is already possible. For example, users can disable specific signals like keystroke capture during onboarding or later from the settings screen. Right now, if those settings are enabled, sensitive data like passwords or private chats can still be captured and sent to the cloud LLM layer (hosted on Vertex AI with no training and no log retention). Even if nothing is stored, I completely understand why that still feels uncomfortable for many users. That’s also why I added support for local LLMs. In that mode, everything becomes fully local. Even though users can manually control capture behavior today, the thing I want most is smarter automatic filtering with strong default protections built in from the start. And honestly, the continuity between my Claude Code and Codex workflows changed dramatically once memory started persisting across them.

Report

3mo ago

Congrats, @screenest_ai. Persistent memory for agents is the piece nobody's solved cleanly yet. Most workflows still require you to re-explain the whole codebase every new session. The passive capture approach makes sense, but I keep thinking about what gets indexed. If it's pulling from all your open files in the background, there has to be some kind of data boundary. Does the memory stay local, or does it leave the machine at any point?

Report

3mo ago

Contextberg

Maker

@ayushi18 The capture granularity is essentially everything. Preventing tracking fundamentally depends on the user intentionally turning signals off when they want privacy. That said, I do think the roadmap needs stronger automatic opt-out mechanisms — for example excluding password fields, specific windows/apps, or sensitive embedded files from capture entirely. After raw data is processed by the LLM, nothing is retained on the cloud side. The memory stays stored locally on the user’s machine. For teams, I can imagine there being demand for centralized/shared memory management in the future. But because this system can contain highly sensitive information, I’m currently very committed to a local-only architecture.

Report

3mo ago

Would this work with hyperagent by air table?

Report

3mo ago

Contextberg

Maker

@oliver_olibrice I researched it a bit more, and it looks like Hyperagent can act as an MCP client.

So if that’s the case, I can pass basically all memory/context through MCP.

The only things I wouldn’t receive back from the Hyperagent side would be browser history, keystrokes, and screenshots.

Report

3mo ago

Cosmic

Persistent context across agent sessions is one of those things that sounds simple but completely changes what's possible, it's the difference between a smart tool and an actual teammate. The MCP approach is the right bet; curious how you're thinking about privacy boundaries for teams vs. individual devs.

Report

3mo ago

Contextberg

Maker

@tonyspiro I think it becomes a completely different thing once sessions persist across different agents — and once you layer the user’s actual work history on top of that. I also want to expand this toward team use cases over time. My idea is to separate it into privacy grades. Level 1: aggregate session-level memory into reusable skills. Level 2: integrate app usage history to systematize how top performers work. Level 3: integrate screen history to evaluate and model work patterns across all users. I expect Level 3 will face the most resistance, so I’m planning to build it gradually starting from Level 1.

Report

3mo ago

Cosmic

@screenest_ai The three-level privacy grade model is exactly right, and the gradient approach is smart. Level 3 resistance is predictable but I'd bet the teams that opt in early will have a compounding advantage that's hard to explain to the ones who didn't. The interesting design question at Level 2 is whether you systematize from the top performers down or let the system discover patterns bottom-up. Top-down is faster to ship, bottom-up is more accurate over time. Which way are you leaning?

Report

3mo ago

Contextberg

Maker

@tonyspiro Thanks, this question really helped me think deeper about the design. I’m leaning toward starting top-down, then gradually transitioning into bottom-up discovery. I think every company’s workflows become highly specific, so defining “what good looks like” purely from raw data would be difficult at the beginning. In that situation, it probably makes sense to first define strong patterns top-down, then use bottom-up signals to compare against those patterns, refine them, and sometimes even discover entirely new successful behaviors. I’m also curious how you approach team knowledge management on your side. I’ve been struggling with the incentive design part — getting people to actively systematize and structure knowledge is surprisingly hard.

Report

3mo ago

"your work as AI agent memory over MCP" is exactly the gap. claude/codex/cursor all forget context across sessions. distribution thought: ship a public template "contextberg for indie founders" - prebuilt memory schemas for product specs, customer feedback, marketing posts. plug-and-play for every solo founder using claude.

Report

3mo ago

Mailwarm

congrats on your launch!

Report

3mo ago

Contextberg

Maker

@naimz Aww thank you!! 🥹 I’ve been working on this for a long time, so your support genuinely means a lot.

Report

3mo ago

1 2

Reviews