this is super cool. can i get it to work as effortlessly as a chrome plugin? I use a lot of different models and a lot of specific/vertical ai on different products that I'd love to not have to constantly give context to to be useful.

love where this is headed!

Short answer is yes - just add the MCP server and you're golden. Longer answer is that most chat clients do not allow the MCP server to automatically inject it's own system prompt which means you have to specifically prompt your AI to write to memobase.

Interesting positioning. Cross-tool memory portability is the right direction. A practical trust layer to add: every recalled item should carry provenance + freshness metadata so agents can choose whether to trust, refresh, or ignore memory instead of treating all recalls as equally valid.

Thanks - glad it resonates. Great idea on the freshness that makes total sense. I'll add that for sure

persistent memory across tools is the right problem to solve. the real friction isnt context length, its context continuity -- picking up a claude session tomorrow feeling like you never left. memobase looks solid for the memory layer. the missing piece most people ignore is session state itself: terminal output, working directory, what commands actually ran. memory without replay is just notes.

Yeah, 100% agree. That's one thing I just thought about yesterday also - Every session should be summarised and written to the data store also, that way sessions become portable contexts. There's potentially a case to be made to have full replay, i.e. the plaintext sessions stored but I'm not entirely sure on how much more valuable that is over a summary.

Do you have thoughts, or a take on that?

summaries are probably 80% of the value at 10% of the storage cost. full replay is nice for debugging weird agent behavior but day-to-day you rarely need the raw transcript.

the interesting edge case is when the summary itself becomes the lossy artifact - like, who decides whats important enough to keep? if the model summarises, it might quietly drop the context that would have mattered most next week.

hybrid might be the move: rolling summary plus last N raw turns.

Right - that makes sense. I see it similarly. My assumption though is that as models get better, the likelyhood of the model missing context that matters the most will get lower and lower.

The hybrid approach though is nice, I'll have a think about that and see if that's something I can incoporate into it. Thanks for the feedback, very much appreciated

yeah models are definitely getting better at prioritizing signal over noise. but theres still an interesting edge case -- the context that matters most is often the stuff you didnt know mattered when you said it. like a throwaway comment three sessions ago that turns out to be load-bearing. hybrid at least gives you a fallback when the model makes a bad bet on whats important.

100% - I'll implement and add that now :)

nice, ship it fast and break things. curious what your persistence layer looks like -- are you serializing full context or just surfacing semantic anchors? the anchor approach is way more interesting imo, lets you reconstruct intent without dragging 50k tokens of cruft into every new session. thats the real trick nobody talks about

I'm not gonna lie - at the moment it's pretty basic as a chunked semantic store where relevant chunks are retrieved in conjunction with some criteria the Agent can pass to the MCP server.

Context usage is definitely a problem that's on my mind also and semantic anchors are one area I'm exploring but don't have a clear architecture for it jotted down yet. The real problem I'm facing right now is how to inject this into say claude or chatgpt and have those agents default use it as a memory layer

for the injection problem - system prompt is your friend. just hardcode "before responding, query your memory MCP tool with relevant keywords" into the system prompt. crude but it works. claude especially follows tool-use instructions pretty reliably when theyre explicit. chatgpt less so honestly. the real trick is making the retrieval call feel like the agents idea, not an obligation - framing it as "you have a memory layer, use it" vs "you must call this tool" makes a suprising difference

That is what I do at the moment - I gotta update the instructions on the website to prod users to set things up this way.

The issue I perceive though is that adding an MCP server alone is not enough to modify the system prompt of the AI Agent. I tried to have the mcp server description be an injected prompt to add these instructions to the system prompt but that doesn't seem to work, I tried adding sampling to the MCP server which supposedly should be able to plug into messages without luck, tried to optimise for chatGPT with an OpenAPI spec etc.

The only way I found that I can get those clients to use my memory layer is by doing what you describe - which is not necessarily the most user friendly/one-click setup I desire

yeah the system prompt injection problem is a real pain. the mcp spec wasnt really designed with "persistent context shaping" in mind, its more request/response than ambient influence.

one hacky workaround ive seen work: expose a tool thats always called first via a resource that hints at it, basically tricking the agent into self-priming. ugly but functional.

honestly though for true one-click youre probably stuck until clients expose better hooks. the spec needs to catch up to the use case.

Tried the resource trick - it's also quite hit and miss. Not all clients properly support resources yet.

I believe you're right and I'm just gonna have to wait or try to engage directly with MCP/Anthropic/OpenAi and further this agenda

Claude Code just added http hooks which are fantastic addition for this problem :) So I've hooked that up with my system. Pretty good start tbh.

http hooks are slick! What’s your anchor strategy? Full context or just the good stuff?

At the moment I'm doing a hybrid but that's still something that needs a lot of trial and error to get right. Doing turn-based context with auto compacting at the moment, but I'm sure that'll change a few more times before it works well

turn-based with auto-compacting sounds tricky. anchors over full context? def way to go for flexibility. keep iterating!

Building in the same space (Maximem Vity, cross-LLM memory as a Chrome extension and OpenClaw plugin). The SMTP analogy really resonates with me. I wrote a comparison a few weeks ago of how ChatGPT, Claude, and OpenClaw actually implement memory under the hood [1] and the architectures are so different that interop feels almost accidental when it works. One thing I keep running into: the hard part isn't storage or retrieval. It's qualification. Deciding what's worth remembering from a conversation vs. what's throwaway context. ChatGPT takes the "summarize the last 15 chats" approach, Claude does on-demand search, and both have real failure modes. We went with a semantic graph that tries to capture relationships between memories (your preference for serverless connects to your AWS project connects to your cost constraints) rather than flat key-value pairs. Still iterating on it honestly. Re: Claude's import-memory launch yesterday, I think the timing validates the category but the approach is fundamentally migration, not sync. You import once, and from that moment your contexts diverge again. Anyone using 3+ tools daily (which is basically everyone I talk to) is back to fragmented memory within a week. Curious about your retrieval approach. Are you doing hybrid search or pure semantic? [1] https://www.maximem.ai/blog/ai-apps-memory

Nice - good to see activity in this space, the more folks are on it the quicker we'll get to a world with a satisfactory solution.

I like your approach - I went down a bit more of a swiss-army-knife one with a hybrid.

"Session memory" where turns are stored in full so that the full context can be retrieved if an important fact is missed in compaction. After a certain time sessions are compacted. I'm currently evaluating a model where only a certain number of turns per session are kept plain-text and compaction happens as a sliding window "knowledge memory" All info is periodically fed into a knowledge graph extraction and a knowledge graph is built and indexed "memory chunking" chunks of memory are stored individually into a vector space where they can be retrieved through similarity search as well as standard semantic searches.

My theory is that giving the agent flexibiity to query the tool best suited for it's current needs is the way to go, as agents/llms become better, they'll only get better at summarisation and tool choice.

Couple reasons I like storing more data over less is that storage is cheap and if compaction misses details, you can prompt your agent to go back to that session you had about X last week and extract that detail.

Similar to you however, it's still also very early stages and the ground is shifting rapidly. Claude Codes HTTP hooks are a great step towards a model that I believe can work assuming it gets adopted as a standard. A big problem of having memory be portable is that you gotta be able to plug it into any client you want, and for the time being, MCP + system prompt is the only option - and even that one's more than flaky.

Would be awesome to chat to someone from OpenAI or Anthropic on their take and pow-wow on options

Update from two days later: we shipped the Claude Code HTTP hooks integration.

The injection problem I mentioned earlier, of memory not loading automatically without system prompt config is now partially solved for Claude Code. SessionStart hook auto-injects memory capabilities, TaskCompleted tracks turns, SessionEnd persists everything.

It's also clarified our thinking on the protocol question: lifecycle hooks (SessionStart, SessionEnd, ContextCompaction) feel like the right abstraction for ambient memory. If other tools did the same, you'd configure memory once and it would work everywhere.

Any thoughts or feedback on the topic are welcome