Agentic memory

One API for capture, search, grounded answers with sources, and the connections between them, across text, PDFs, images and video. Usage-based pricing.

Get your API key Quickstart API reference

# put something into memory
curl -X POST https://api.fryri.com/v1/capture \
  -H "Authorization: Bearer fryri_sk_..." \
  -H "Content-Type: application/json" \
  -d '{"text": "Lease renewal due March 1, landlord wants an answer by Friday",
       "end_user_id": "user_42"}'

# search it later
curl "https://api.fryri.com/v1/search?q=lease+due&end_user_id=user_42" \
  -H "Authorization: Bearer fryri_sk_..."

Memory

A memory that takes a note, a PDF, an image or a video and keeps it ready to recall.

Search

Ask anything and get the answer back, from a fuzzy idea to an exact phrase to how the people and things inside connect.

Answers

One call returns a grounded answer with its sources, on any model you choose.

Webhooks

A callback fires the moment something is ready, so nothing has to sit and wait.

Research goes in too

Research

Kick off a research run and get back a sourced comparison table or a profile of a person or company. The result lands in the memory, so the next question recalls it instead of researching again.

Scope it however you want

Pass end_user_id on any call to keep a separate, private memory under that id, or leave it off for one shared memory. Each id stays walled off from the rest, and one call removes it completely.

Pay only for what you use, and cap any key so a bill can never surprise you.
Drop it into an agent over MCP, or into your app with the Python and TypeScript clients.
Get a key, add credit and see your spend from the developer console.