Smara -- One Memory for All Your AI Tools

Persistent, cross-platform memory for AI agents. Ebbinghaus decay scoring, contradiction detection, source tagging, and vector search. Works with Claude Code, Cursor, Codex, and any tool that speaks MCP or REST.

---

The Problem

Every AI tool has its own isolated memory. Tell Claude Code you prefer Python -- switch to Cursor, it has no idea. Open Codex -- starts from scratch.

Smara fixes this. One memory pool, shared across every AI tool. Memories are ranked by semantic relevance and Ebbinghaus decay -- recent, frequently accessed facts stay strong while old trivia fades naturally.

---

Quick Start

1. Get an API key

curl -X POST https://api.smara.io/v1/signup \
  -H "Content-Type: application/json" \
  -d '{"email": "you@example.com"}'

{
  "api_key": "smara_...",
  "message": "Save this API key -- it cannot be recovered."
}

Or sign up at smara.io.

2. Add to your AI tools

Claude Code, Cursor, Windsurf (MCP):

Add to your MCP config (~/.claude/mcp_config.json, .cursor/mcp.json, etc.):

{
  "smara": {
    "command": "npx",
    "args": ["-y", "@smara/mcp-server"],
    "env": {
      "SMARA_API_KEY": "smara_your_key_here"
    }
  }
}

Restart your tool. Memory is automatic -- context loads at conversation start, new facts are stored silently.

Any tool (REST API):

Three endpoints. That's it.

---

API Reference

All /v1/* endpoints require Authorization: Bearer <API_KEY>.

Store a memory

curl -X POST https://api.smara.io/v1/memories \
  -H "Authorization: Bearer $SMARA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "user_42",
    "fact": "Prefers Python over TypeScript for backend work",
    "importance": 0.8,
    "source": "claude-code",
    "namespace": "default"
  }'

{
  "action": "stored",
  "id": "a1b2c3d4-...",
  "source": "claude-code",
  "namespace": "default"
}

Field	Type	Required	Description
user_id	string	Yes	Your user identifier
fact	string	Yes	The fact to store (1-2000 chars)
importance	number	No	0.0-1.0, default 0.5
source	string	No	Which tool stored this (e.g., "claude-code", "cursor")
namespace	string	No	Partition memories (default: "default")

Smart storage:

• Duplicate (cosine >= 0.985): returns existing memory, no duplicate created
• Contradiction (cosine 0.94-0.985): soft-deletes old fact, stores new one
• Returns "action": "stored", "duplicate", or "replaced"

Search memories

curl "https://api.smara.io/v1/memories/search?user_id=user_42&q=editor+preferences&limit=5" \
  -H "Authorization: Bearer $SMARA_API_KEY"

{
  "results": [
    {
      "id": "a1b2c3d4-...",
      "fact": "Prefers Python over TypeScript for backend work",
      "importance": 0.8,
      "decay_score": 0.97,
      "similarity": 0.88,
      "score": 0.91,
      "source": "claude-code",
      "namespace": "default",
      "created_at": "2026-03-30T12:00:00Z"
    }
  ]
}

Param	Type	Required	Description
user_id	string	Yes	Your user identifier
q	string	Yes	Natural language search query
limit	number	No	1-50, default 10
source	string	No	Filter by source tool
namespace	string	No	Filter by namespace

Get user context

Pre-formatted context for LLM system prompts. Works with or without a query.

# With query (semantic search + decay)
curl "https://api.smara.io/v1/users/user_42/context?q=preferences&top_n=5" \
  -H "Authorization: Bearer $SMARA_API_KEY"

# Without query (top memories by decay score)
curl "https://api.smara.io/v1/users/user_42/context?top_n=5" \
  -H "Authorization: Bearer $SMARA_API_KEY"

{
  "context": "[1] (importance: 0.8, decay: 0.97, source: claude-code) Prefers Python over TypeScript...",
  "memories": [...]
}

Delete a memory

curl -X DELETE https://api.smara.io/v1/memories/<memory_id> \
  -H "Authorization: Bearer $SMARA_API_KEY"

Returns 204 No Content on success.

Check usage

curl https://api.smara.io/v1/usage \
  -H "Authorization: Bearer $SMARA_API_KEY"

{
  "plan": "free",
  "memory_limit": 10000,
  "memories_used": 250,
  "memories_remaining": 9750
}

---

How It Works

Ebbinghaus Decay

Every memory's relevance decays over time using the Ebbinghaus forgetting curve. High-importance memories decay slowly; low-importance ones fade fast. Memories that are frequently accessed get reinforced.

Blended Ranking

Search results are ranked by a blend of semantic similarity and temporal decay. Relevant + recent beats relevant + old.

Cross-Platform Source Tagging

Every memory is tagged with its source tool. A fact stored via Claude Code is instantly available in Cursor, Codex, or any connected tool. Filter by source or see everything.

Namespace Isolation

Partition memories by namespace (e.g., "work", "personal", "test"). Deduplication and contradiction detection are scoped per namespace.

---

MCP Server

The MCP server (@smara/mcp-server) provides automatic memory for MCP-compatible tools.

npx -y @smara/mcp-server

Environment variables:

Variable	Required	Description
SMARA_API_KEY	Yes	Your API key
SMARA_API_URL	No	API URL (default: https://api.smara.io)
SMARA_USER_ID	No	Default user ID for all operations
SMARA_SOURCE	No	Source tag (default: "mcp")
SMARA_NAMESPACE	No	Default namespace

Auto-memory behavior:

1. At conversation start: loads stored context automatically
2. During conversation: silently stores new facts
3. On "remember this" / "forget that": explicit store/delete

See @smara/mcp-server on npm.

---

Integrations

Tool	Method	Status
Claude Code	MCP server	Live
Cursor	MCP server	Live
Windsurf	MCP server	Live
CrewAI	Python SDK	smara-crewai
LangChain	Python SDK	smara-langchain
Paperclip	Plugin	paperclip-plugin
OpenAI / Codex	REST API	Use endpoints directly

---

Self-Hosting

Docker (recommended)

git clone https://github.com/smara-io/api.git
cd api
VOYAGE_API_KEY=your-key docker compose up -d

API runs on localhost:3011, Postgres on localhost:5433.

Manual

Requirements: Node.js 20+, PostgreSQL 15+ with pgvector, Voyage AI API key.

git clone https://github.com/smara-io/api.git
cd api
npm install
npm run build

# Set environment variables
export DATABASE_URL=postgresql://user:pass@localhost:5432/smara
export VOYAGE_API_KEY=your-key

npm start

The API auto-migrates on startup (creates tables, indexes, extensions).

---

Pricing

Plan	Memories	Price
Free	10,000	$0
Developer	200,000	$19/mo
Pro	2,000,000 + $0.50/10K overage	$99/mo

Sign up free -- no credit card required.

---

Links

• Website: smara.io
• API Docs: api.smara.io/docs
• MCP Server: npm @smara/mcp-server
• Blog: How Ebbinghaus Forgetting Curves Make AI Agents Smarter
• Twitter: @SmaraMemo

---

License

MIT