Supermemory MCP is live · See the launch

The universal memory API for the AI era

Stop building retrieval from scratch. Personalise LLMs for your users. Built for developers who ship.

Get supermemory

Github (9k+ stars)Docs

Trusted by Open Source, enterprise, and more than 35,000 of you

Starred by over 9,000 users on Github

Flow uses supermemory to build the cursor for writing.

MedtechVendors uses supermemory to search through 500k vendors. ... and more

Context is everything

Without it, even the smartest AI is just an expensive chatbot

developer-memory-hell.terminal

~ memory.engineering

$ init vector_database

✗Way too expensive. Time to switch.

✗Painfully slow. Let's try another.

✗Won't scale. Back to square one.

✗Maintenance nightmare. Need alternatives.

$ choose embedding_model

⚠Which model fits your use case?

⚠Confusing performance tradeoffs

⚠Can't keep up with new releases

$ setup connection_sync

✗Sync failures between data sources

✗API rate limits during large syncs

✗Authentication tokens expire constantly

$ calculate scaling_costs

⚠Costs explode at production scale

⚠Performance degrades as data grows

⚠Engineering hours pile up fast

$ handle format_parsing

✗Markdown: Tables break everything

✗HTML: Scripts and styles interfere

✗PDF: Layout ruins extraction

✗Word docs: Unpredictable formatting

$ init multimodal_support

✗Websites: JS & rate limits are messy

✗PDFs: OCR fails, extraction inconsistent

✗Images: Need vision models now?

✗Audio/Video: Transcription costs soar

BREAKING NEWS

New retrieval strategy just dropped:

Hierarchical Enhanced Retrievers with Dynamic Adjustable Query Expansion

(HERDAQE)

$ compare retrieval_methods

RAG

CAG

HYDE

HERDAQE

AAAAAAAAAAAAAAAAAAAAAAAAA

Error: Memory infrastructure failed to initialize. Time to fix: 6-8 weeks.

Just use Supermemory

One API call. All formats. Any source. Always bleeding edge. Done.

import supermemory from 'supermemory'

// That's it. Ship your product.

Start Building →

Interoperability

Model-agnostic APIs.

Supermemory works with any LLM provider. So you can keep switching, without lock-in.

Performance

Sub-400ms latency at scale.

Supermemory is built for speed and scale. We re-imagined RAG to be faster and more efficient.

Tooling

Works with AI SDK, Langchain, and more.

Just one line of code to get started. Yes. you heard that right.

Language Agnostic

Deploy in a day, not months.

SDKs available for Python and Javascript.Learn more