A knowledge base your chatbot actually remembers.
Connect any chatbot or AI system to Memocore over a simple API. Store what you know once, then recall only the memos that matter for each message — instead of stuffing your entire knowledge base into every prompt.
REST & MCP · semantic search · cut token costs
Prompt-stuffing is expensive — and it doesn’t scale.
Pasting your whole knowledge base into every request burns tokens, fills the context window, and still goes stale.
Send a few memos, not the base
Fetch only what the question actually needs, and the token bill drops with it.
The window stays free
Small, relevant context leaves room for the conversation and keeps answers sharp.
Always the current answer
Change a memo and the answer changes with it. No retraining, no re-indexing pipeline to run.
Store once. Recall only what’s relevant.
Three steps from an empty project to a bot that answers from your own memory.
Fill your project
Add everything your bot should know to a Memocore project — write memos yourself or let your connected assistants save them as you work.
Create an API key
Issue a key for your company and tick the projects it may reach. Everything outside them stays out of its reach.
Connect it to your bot
Point your chatbot at the API so it can search Memocore for the memos a message needs — and create new ones as it learns.
Everything the API gives you.
Four endpoints and one bearer token — the same search your team uses inside Memocore, available to your own code.
Search that understands the question
Query with the user’s message and get back the handful of memos that answer it. Meaning and keywords are searched together and the two rankings are fused, so a memo comes back even when it shares no words with the question. Need a plain keyword match instead? Turn the vector side off with one parameter.
Create memos from your app
Your product doesn’t only read. Post a memo and it lands in the project you chose, with a title, tags and a short preview.
Filter by tags
Require the tags a memo must carry, exclude the ones it must not, and list every tag your key can reach.
Keys scoped to projects
A key reaches only the projects you tick for it, so the bot answering customers never touches internal projects.
100 requests per second
That’s the limit per key, so a busy chat doesn’t queue behind a nightly import. Issue as many keys as you need.
The whole surface — four calls
No SDK to install and no schema to design. A bearer token and these endpoints are the entire integration.
Common questions
Nothing on top of your plan. API keys come with the team plan — $15 per active seat per month, billed through Stripe — and that covers unlimited keys, unlimited memos and projects, and the search behind them. There is no separate per-request price.
Yes. The API both searches a project and creates memos in it, so your product can save what it learns and recall it later. Each key is limited to the projects you select for it.
As many keys as you need — one per bot or environment is the usual pattern. Each key is limited to 100 requests per second and to the projects you tick for it.
No. Semantic search is built in — there’s no embeddings pipeline or vector store for you to run and maintain.
Instead of sending your whole knowledge base on every call, you retrieve only the few memos relevant to the message — a few hundred tokens instead of thousands.
Any chatbot, agent or AI system that can make an HTTP request, plus any client that speaks the Model Context Protocol.
Keys stop working while the team plan is inactive, and calls come back with a clear error instead of stale data. Your memos stay where they are, and you can export everything at any point.
Still wondering about something? Ask the chatbot in the bottom-right corner — it knows Memocore inside out.
Give your AI systems a memory.
Create your account with just your email, fill a project, then issue a key and point your bot at it.
API keys come with the team plan · REST & MCP · semantic search