Do I need to change my code?
Do I need to change my code?
Only the base URL, a
boundary_id query parameter, and — in BYOK mode — one header. Your prompts, streaming, tool calls, and response handling stay identical, because Memory Router speaks the same API as your provider.What is a Boundary and do I always need one?
What is a Boundary and do I always need one?
A Boundary is the memory scope for a conversation — it binds requests to a workspace, project(s), and actors so the Router knows whose memory to read. You create it once in the console and pass its id as
?boundary_id= on each call. Requests without a boundary_id still work, but pass through with memory disabled.Which providers and models are supported?
Which providers and models are supported?
OpenAI (Chat Completions, Responses, Images) and Anthropic (Messages) protocols. In hosted mode, GPT, o-series, and Claude models — plus image models such as
gpt-image-1 — are built in and callable by their native names; the console’s Available Models page lists them all. No embedding models are available yet. See Deployment Modes for the current matrix.Can I generate images?
Can I generate images?
Yes. Call image models such as
gpt-image-1 through the OpenAI Images endpoint (images.generate in the OpenAI SDKs), with the same base URL and key as chat models. Image generation does not use memory. See Image Generation.What is the difference between BYOK and MemoryLake-hosted?
What is the difference between BYOK and MemoryLake-hosted?
BYOK (Bring Your Own Key) means you supply your own provider key plus a MemoryLake key — model billing and rate limits stay with your provider account, while memory is billed to your MemoryLake balance. MemoryLake-hosted needs only a MemoryLake key: we run the major models for you, so you skip provider sign-up entirely. Both modes use the same endpoints; the difference is which keys you send.
In BYOK mode, is my provider key safe?
In BYOK mode, is my provider key safe?
Yes. Your provider key is encrypted in transit and forwarded to the provider on each call. MemoryLake never stores, logs, or reuses it — the design is stateless, per-request passthrough with zero server-side key storage.
Does memory add latency to my calls?
Does memory add latency to my calls?
Memory recall happens before the model call and is optimized for low overhead. The Router does not write memories, so nothing extra runs after the response.
How do I get access?
How do I get access?
Sign in to the console, create an API key, and open Integrations → Memory Router API to create a Boundary and copy ready-made snippets. The Quickstart walks through it.