Tanka
Team Messenger with AI Smart Reply & Long-Term Memory
4.4•11 reviews•2.6K followers
Team Messenger with AI Smart Reply & Long-Term Memory
4.4•11 reviews•2.6K followers
🚀 Meet Tanka, the AI MESSENGER with MEMORY for TEAMS. Integrated with WhatsApp, Slack, Telegram & Gmail, Tanka learns from your past chats to deliver context-aware AI Smart Replies, to-dos & insights as your team's second brain & personalized memory bank! 🧠
Shram
I love the idea of an AI that actually remembers past conversations. That’s such a huge leap from the usual chat tools that just respond in the moment and forget everything. Tanka is a great tool for team communication. I can see how it’d be super helpful for onboarding new team members too, having all that knowledge readily available would make things so much easier. Excited to see how this evolves!
Congrats on the launch!
Best wishes and sending lots of wins to the team :) @kissonlin
Tanka
@whatshivamdo Exactly! We built the product because we had such pain. Thank you for liking it! Hope you enjoy using the product :) Also - good luck with your launch too!
Tanka
@whatshivamdo exactly, and the cost of forgetting, or even mis-remembering, something could be huge. pls let us know if Tanka works for you!
Can't tell you the no.of times I've had to search/scroll through Slack to find past messages & convos...
...if this works as smoothly as it sounds, it would give me a massive break :)
Might have to convince the team to use it too - Best of luck w/ the launch Kisson & the rest of the team!!
Tanka
@cranqnow Thanks Sam. Let's chat! Would love to get your feedback and follow-up on your experience :)
Tanka
@cranqnow Keen to know if we are up to your expectation! Let me know any feedback, we'd love to make it useful for you :) Thank you!!
I'll be sure to let you know Kisson!!
@kissonlin @alanchristophx @atzhuyun Really interesting re: long term memory - We've got a similar tool (on the AI brain side) on our end but its for creators vs internal repo - I am curious - lets presume I don't want AI to be trained on certain internal IP - how are you handling that use case?
Also curious how you're handling your internal prompting to keep things are vague but comprehensive as possible for users?!
Anyhoo killer product and congrats again on the rollout!
Tanka
@alanchristophx @atzhuyun @dzaitzow Thanks for your comment! An AI brain for creators sounds really cool, I'd love to use if myself.
We don't train our model with user data. Currently, all memory and chat history is encrypted and stored on AWS cloud in the US.
In the future, we'll provide privately hosted version for enterprises for larger scale hosting requests.
On the prompting side, we actually went way beyond just data retrieval. We built memory graph for users. Its abstract understanding and causal inference capability enable Tanka to keep things comprehensive. It can even understand some implicit context.
We are just 6 months into building this product. More to come very soon!
Also - good luck with your launch!
Sounds really interesting! Feel free to ping me if you're interested in checking out the product (even before launch)
Congrats on your launch, Team Tanka!
This looks fantastic! How does Tanka ensure sensitive data security, especially when integrated with Gmail?
Tanka
@bonvisions Hi Bon yes security and compliance is critical for us. Tanka adheres to industry-leading security frameworks, including SOC 2 and ISO 27001. All conversations are encrypted using AES-256 encryption at rest and TLS 1.2+ in transit, ensuring that your data remains protected at all times. Tanka does not access, monitor, or retain your conversations for AI training or model improvement. Your data is only yours and is never used to train any AI models. We also ensure strict data isolation—your information is never mixed with data from other users or organizations. Only authorized personnel, under strict access control, can handle sensitive information. Our security infrastructure undergoes regular penetration testing, real-time monitoring, and strict access control to safeguard your data.
Let's chat more about your use cases or potential concerns!
Congrats! It looks great, I'm sure I'll give it a try, especially seeing your beautifully designed product page. Keep it up! ❤️❤️❤️
Tanka
@cloudkoonly Thanks for the kind words, please let us know your feedback!
NYX
Congrats on such a great launch!! I think our CSAT team can use this to automate most queries with personalized responses. Can we implement it organization-wide? Or will each individual have to use it individually?
Tanka
@shikhar_agrawal3 Thank you! This is really good to know. Yes! Tanka is made for organizations. I'll DM you for collaboration.
Also - good luck with your upcoming launch!
AFFiNE
Looks like a great concept. Wondering what's the available context size. Does it have a light memory of all topics that it can dive into? Wondering if it forgets old topics or keeps some type of archive reference.
Tanka
@shortcipher5 Thank you! Great question!
Context window is exactly the challenge LLMs face. And with RAG, the problem still exist because complex questions require a lot of context and connecting the dots. Memory is useful in this context because for every xxxxx conversations we chunk the data and structure the memory into storylines with hierarchy, so later on when retrieval is needed memory serves as the index of a book, solving context issues :)
For now it's quite accurate. We still continuously improve the product so to make it more accurate after months, years of storage.