Your call center remembers every caller. Cross-call. Per-caller isolated. Two API calls.
Your chatbot remembers every user. Cross-session. Per-user isolated. Two API calls.
Your voice agent remembers every caller — across every call. The second call picks up exactly where the first ended: who they are, what's on file, what they came back for.
Two users. One chatbot. Watch each user's memory persist across sessions — and watch the wall between them hold. Sarah can never see Mike's data. Mike can never see Sarah's.
Your dashboard. Every user's context. Complete isolation.
PraqtorDB sits behind whatever voice stack you run. Add caller memory with two HTTP calls — no SDK lock-in, no platform migration.
Wire PraqtorDB into your Retell agent's webhook. Recall the caller's history before the agent speaks, save every turn after.
Native Vapi integration is on the way. Per-caller memory and recall, dropped straight into your assistant flow.
Bland support is in progress. Give every Bland call a persistent, per-caller memory across conversations.
LiveKit Agents integration is coming. Stream-time recall and capture for real-time voice pipelines.
ElevenLabs conversational agents are next up. Remember who's calling and what they came back for.
Any stack that can make an HTTP request. If your voice agent speaks REST, it has caller memory today.
PraqtorDB for Apps sits behind whatever you already run. Add memory with two HTTP calls — no SDK lock-in, no platform migration.
Drop the two calls into your bot's turn handler. Works with the Bot Framework SDK out of the box.
Call PraqtorDB from a webhook fulfillment. Save the turn, search prior context, return a grounded reply.
Wire it into a Lambda code hook. Each Lex intent reads and writes the user's isolated memory.
Any stack that can make an HTTP request. Python, Node, Go — if it speaks REST, it has memory.
Front-end for GPT, Claude, Gemini, Llama. Search memory, inject it into your prompt, save the reply.
LangGraph, AutoGen, the OpenAI Agents SDK. Give every agent run a persistent, per-user memory.
The same engine that powers PraqtorDB — re-pointed at your application's end users. Each user gets their own isolated memory, knowledge graph, and AI-generated profile.
Users pick up exactly where they left off. Context survives across sessions, devices, and weeks of silence.
Sarah can never see Mike's data. Every memory is scoped to a single user_id. Hard walls, enforced by the engine.
Entities and relationships are extracted automatically from every conversation. No tagging, no schema design.
AI-generated user profiles distilled from each user's full history — preferences, intent, and context at a glance.
Chat, files, documents. Ingest from anywhere your users interact and it all flows into one per-user memory.
POST /v1/memory to save. POST /v1/search to retrieve. That's the whole integration.
Create. Send. Extract. Search. Your chatbot gains memory in four steps — and most of them happen automatically.
Register your app on PraqtorDB for Apps and get an API key. One namespace for all your end users.
Send each turn via the API — just two calls. Tag every message with the end user's user_id.
The engine builds entities, notes, and deep memory per user — automatically, in the background, isolated per user_id.
Your chatbot searches that user's memory before responding. Grounded answers, never crossing the user boundary.
# 1. SAVE — store this turn in the user's isolated memory requests.post("https://api.praqtordb.com/v1/memory", headers={"Authorization": "Bearer $PRAQTOR_API_KEY"}, json={ "user_id": "sarah", # per-user isolation "role": "user", "text": "I want to buy cheese, bread, and yogurt" }) # 2. SEARCH — retrieve this user's memory before you reply ctx = requests.post("https://api.praqtordb.com/v1/search", headers={"Authorization": "Bearer $PRAQTOR_API_KEY"}, json={ "user_id": "sarah", # never returns Mike's data "query": "what did I want to buy?" }).json() # -> "cheese, bread, and yogurt" (Sarah's memory only)
Every feature included at every tier. Start free for three months, then pick a plan based on your user count.