AI Context Flow
Reusable AI Memory for Smarter Prompts Anywhere
Start new thread

AI memory quickly becomes a junk drawer. We built the organization layer.

by•

You can now store unlimited context for AI, but without structure, your memories quickly become a messy pile of information that you have no idea how to work with. We've been there too.

So, we've been heads-down building, and Memory Studio now has a Notion-style editor: with version history, rich formatting, enhanced file understanding, markdown support and more.

Here's what's we shipped:

🗒️ Memory Studio - completely reworked

Think Notion-style editing, but living inside your AI memory layer.

Working with your memories is now a proper experience:

→ Full Notion-style document editor inside every memory item
→ Rich-text ↔ Markdown toggle
→ Auto-convert PDFs, Word, PowerPoint, Excel & HTML to clean Markdown on upload
→ AI Vision 👀 — photos, screenshots & scanned PDFs are read and described by AI
→ On-demand AI Format with upfront token estimate
→ Smarter summaries drawn from beginning, middle & end of your documents
→ Free text edits — no tokens, no AI, just fast editing
→ Multi-file drag-and-drop with live progress
→ Cleaner saved chat transcripts with User/Assistant labels

🔌 Browser Extension: Override

The prompt optimizer also got an upgrade. An agent sitting with you inside other agents, that you can give directions to.

Type <instruction> in any prompt to take direct control over what gets pulled from your memory. No more guessing - just tell it exactly what you need.

Examples:
▸ write release notes use my exact changelog → pulls it verbatim.
▸ draft a reply include my standard pricing table → attaches it exactly.
▸ set up the project use my saved env config → injects it as-is.


We're building toward a world where every AI tool you use has the right context automatically. And you can easily audit and correct the context your multi-agent workflows are getting.
This is another step toward that.


If you've tried AI Context Flow, we'd love to hear what you think. Drop a comment or leave us a review on the Chrome Store 🙏

33 views

Add a comment

Replies

Best

Strong direction. AI memory only becomes useful when structure is as good as storage , otherwise it just turns into scalae clutter.

The override is interesting . How granular can the instructions get? Can you target specific memories projects or rtime ranges?

yes you can! There’s a complete reasoning model behind it so you can really get creative with what to fetch and how

the junk drawer problem is real and the override is the most interesting part of this to me. most memory layers assume the retrieval will just figure it out, but giving the user an inline escape hatch to say "use this exactly" is the kind of control that actually builds trust. curious how it handles conflicts when the injected context contradicts what the memory layer would have pulled automatically. also thinking about multi-agent workflows where the agent issuing the prompt isn't a human typing but another agent. does the override syntax work when the prompt is generated programmatically, or is it really designed for human-in-the-loop use right now?

this works in extension right now, we will add this to mcp soon, then it will become proper agent to agent without human in the loop. Honestly this is one of the creative experiments we have deployed and will hone it more based on feedback. If the agent would have fetched something else but aicf override asks something else, the aicf override has priority.

 > without human in the loop

tbh ...thats scary !!!