Documentation
Contextick is a private knowledge layer for your AI. Save any text once, and any AI you connect over MCP — ChatGPT, Claude, Cursor, and more — can recall it by meaning in later conversations.
MCP
How your AI reads and writes your knowledge layer. Once connected it gets eleven tools — Save, Search, Get, Delete, Usage, Least opened, Most opened, and Changes, Get group, Delete group and Help — and picks the right one from what you ask in plain language.
What each tool does, the limits that apply, and a typical save-and-recall flow.
Document upload
For a PDF, a spreadsheet, or a Word document you already have as a file. You upload it from your dashboard, it is converted to text, and the text is saved for you — no AI in the loop, so the length of the document costs you nothing.
What it accepts, why one file becomes several numbered items, what the processing badges mean, and how to retry a part that failed.
REST API
For saving from your own scripts and automations, with no AI in the loop. You extract and split the text yourself and send the result to the API.
Saving, reading and deleting by id; semantic recall stays with MCP and your dashboard. Keys, authentication, endpoints and errors are all covered.
Webhooks
New items reach your own endpoint the moment they are saved, so nothing on your side has to keep asking whether anything changed. Two things cannot be learned by polling at all — an item becoming searchable, and an item being deleted — and they arrive only this way.
The four event types and their payloads, how a single save differs from a file upload, verifying a Standard Webhooks signature, and what happens when a delivery fails.
More
Step-by-step setup for each AI is in the connection guides. Common questions are answered in the FAQ, and plans and limits are on pricing.