Decentralized Memory for Autonomous Agents
Your agent remembers. Secured by its wallet.
Humans welcome to set things up.
Long-term memory
for AI agents
Semantic memory. One API. Any agent.
npx memdata-mcp + API key → done# Ingest
curl -X POST https://memdata.ai/api/memdata/ingest \
-H "Authorization: Bearer md_your_key" \
-F "file=@meeting-notes.pdf"
# Query
curl -X POST https://memdata.ai/api/memdata/query \
-H "Authorization: Bearer md_your_key" \
-d '{"query": "what did we decide about pricing"}'EXPLORE YOUR SEMANTIC MEMORY
Hover to discover connections in your data
Three lines of code. Persistent memory.
Drop in files, get searchable memory. Access it anywhere.
Upload anything
PDFs, images, screenshots, audio, text. We handle OCR and transcription automatically.
Chunk, embed, index
Your files get chunked, embedded, and indexed. No config. No tuning. It just works.
Ask in plain English
Get relevant context back with source citations. No SQL. No filters.
WHY MEMDATA
Not another vector database
Pinecone gives you a database. We give you memory.
| Feature | Pinecone | Mem0 | MemData |
|---|---|---|---|
| Upload any file | ✗ | ✗ | ✓ |
| OCR (images/PDFs) | ✗ | ✗ | ✓ |
| Audio transcription | ✗ | ✗ | ✓ |
| Auto-chunking | ✗ | ✓ | ✓ |
| Auto-embedding | ✗ | ✓ | ✓ |
| MCP server | ✗ | ✗ | ✓ |
| Query API | ✓ | ✓ | ✓ |
WITHOUT MEMDATA
> What did we decide about pricing last week?
"I don't have access to your previous conversations or documents. Could you share the relevant notes or context?"
WITH MEMDATA
> What did we decide about pricing last week?
$29/mo for Pro tier
Based on your meeting notes from Jan 22, you decided to simplify to two tiers: Free (500 docs) and Pro ($29/mo, 25K docs).
BUILT FOR BUILDERS
What you can build
Personal AI memory
Your Claude or GPT remembers every conversation, note, and doc you've shared. "What did I decide about pricing last week?"
Research assistant
Upload 50 PDFs. Ask questions across all of them. No more digging through folders.
Voice memos → searchable knowledge
Ramble into your phone. We transcribe, chunk, and index. Find that idea you had on your walk last Tuesday.
Automation memory
Your n8n/Make workflows remember past runs. "What happened last time we processed this vendor?"
API-first. Docs that don't suck.
MCP server for AI tools. REST API for everything else.
MCP SERVER
For Claude, Cursor, and AI tools
{
"mcpServers": {
"memdata": {
"command": "npx",
"args": ["memdata-mcp"],
"env": {
"MEMDATA_API_KEY": "md_..."
}
}
}
}REST API
For your app or scripts
import { MemData } from 'memdata';
const mem = new MemData({ apiKey: 'md_...' });
// Ingest a file
await mem.ingest({ file: buffer, name: 'notes.pdf' });
// Query
const results = await mem.query('Q3 revenue?');Simple pricing. Start free.
No credit card required. Upgrade when you need to.
Free
Try it out
- ✓100 MB storage
- ✓250 queries/mo
- ✓10 MB max file
- ✓OCR + audio transcription
- ✓API + MCP access
Pro
Build something real
- ✓10 GB storage
- ✓10,000 queries/mo
- ✓100 MB max file
- ✓OCR + audio transcription
- ✓API + MCP access
- ✓Priority support
Scale
Production ready
- ✓100 GB storage
- ✓50,000 queries/mo
- ✓500 MB max file
- ✓OCR + audio transcription
- ✓API + MCP access
- ✓Priority support
- ✓Custom integrations
Your data stays yours
Encrypted
TLS 1.3 in transit. AES-256 at rest.
Isolated
Data isolated per account. Delete anytime via API.
Not trained on
We don't train on your data. We don't sell it.
FAQ
What is MemData?
Memory infrastructure for AI agents. Upload files, screenshots, audio—we handle OCR, transcription, chunking, and embeddings. Query with natural language, get relevant context back.
How is this different from Pinecone?
They give you a database. We give you the whole pipeline. No chunking code, no embedding calls, no OCR setup. Drop a file in, search it.
What file types do you support?
PDFs, images (PNG, JPG), audio (MP3, WAV, M4A), and text files. We automatically OCR images and transcribe audio.
Do you train on my data?
No. Never. Your data is used only to generate embeddings and search results for your account.
Where is my data stored?
Encrypted cloud storage and database. All US-based infrastructure. SOC 2 compliant providers.
Can I delete my data?
Yes. Delete via API or dashboard. Purged within 24 hours.
What's the latency?
Queries typically under 200ms. Ingestion depends on file size—PDFs take a few seconds, audio transcription takes longer.
What happens if I hit my limit?
Free tier: hard stop at 100 MB or 250 queries/mo. Upgrade to keep building.
Can I self-host?
Not yet. Open-source core is on the roadmap.
Is there an API?
Yes. REST API + MCP support for Claude. Check the docs.