> features

everything your ai needs to remember

persistent memory for claude code, cursor, and every ai tool in your stack. local-first. cloud-synced. eu-hosted.

‹ 01 ›

core api

Full CRUD with batch support. Store decisions, patterns, and project context through a clean REST API. Up to 100 memories per batch call.

store a memory

curl -X POST https://api.usememra.com/v1/memories \
  -H "Authorization: Bearer memra_live_..." \
  -H "Content-Type: application/json" \
  -d '{"content": "User prefers dark mode", "type": "preference"}'

‹ 02 ›

semantic recall

Query by meaning, not keywords. Hybrid retrieval combines vector embeddings, entity graphs, and reciprocal rank fusion for the most relevant results.

recall memories

curl -X POST https://api.usememra.com/v1/memories/recall \
  -H "Authorization: Bearer memra_live_..." \
  -H "Content-Type: application/json" \
  -d '{"query": "user preferences"}'

‹ 03 ›

privacy shield

Automatic PII detection and deterministic masking. Emails, phone numbers, and IDs are replaced with consistent pseudonyms before storage. Embedding providers never see raw PII.

regex detection

Automatically detects emails, phone numbers, SSNs, and credit card numbers before storage.

deterministic masking

Same input always produces the same token per tenant, preserving cross-memory recall consistency.

encrypted vault

AES-256-GCM encrypted token storage with unique nonces. Original values recoverable only with scoped auth.

‹ 04 ›

multi-tenant isolation

Isolate memories by tenant ID. Full deletion cascade per namespace -- content, embeddings, PII tokens, cache, audit log. Export everything with one API call.

‹ 05 ›

official sdks

First-party SDKs for PHP/Laravel, Python, and TypeScript. Type-safe, fully tested, with sync and async support.

php / laravel

$memory = Memra::store(
    'User prefers dark mode',
    type: 'preference'
);

python

memory = client.memories.create(
    content="User prefers dark mode",
    type="preference"
)

typescript

const memory = await memra.memories.create({
    content: 'User prefers dark mode',
    type: 'preference'
});

install

composer require memra/laravel-sdk

install

pip install memra-sdk

install

npm install @memra/sdk

‹ 06 ›

mcp server

Connect any AI coding tool to Memra via the Model Context Protocol. 13 tools for storing, searching, syncing, and managing memories -- works with Claude Code, Cursor, Zed, and any MCP-compatible IDE.

‹ 07 ›

local-first mode

Run Memra entirely on your machine. Zero data leaves your laptop. Semantic search powered by local embeddings. Opt into cloud sync only when your team needs shared knowledge.

zero network

Everything runs locally -- flat files, SQLite search index, and sentence-transformer embeddings. No API calls, no cloud dependency.

auto-capture

Install a hook for Claude Code. After every session, decisions and patterns are automatically extracted and saved locally.

team sync

Opt-in cloud sync with PII masking. Share team knowledge across machines. Conflicts detected and preserved for manual resolution.

get started locally

pip install memra-local
memra mcp          # Connect to Claude Code / Cursor / Zed
memra hooks install # Auto-capture session learnings

‹ 08 ›

ready to build smarter?

Local mode is free forever. Cloud plans start at free too.

create free account