Your AI forgets everything between sessions. Deep Recall fixes that. Install the MCP server and your AI agent remembers conversations, preferences, facts — forever.
pip install deeprecall-mcp
That's it. One package, no dependencies to manage.
Sign up at deeprecall.dev with Google, GitHub, or email. You'll get your API key instantly — no email verification needed to start using the API.
Or use the API directly:
curl -X POST https://api.deeprecall.dev/v1/signup \
-H "Content-Type: application/json" \
-d '{"name": "Your Name", "email": "you@example.com", "password": "your-password"}'
Copy the api_key from the response. Save it — it's only shown once. You can always create more in your dashboard.
{
"mcpServers": {
"deeprecall": {
"command": "deeprecall-mcp",
"env": {
"DEEPRECALL_API_KEY": "ec_live_YOUR_KEY"
}
}
}
}
Add this to ~/.claude/settings.json or run /mcp in Claude Code to manage MCP servers.
{
"mcpServers": {
"deeprecall": {
"command": "deeprecall-mcp",
"env": {
"DEEPRECALL_API_KEY": "ec_live_YOUR_KEY"
}
}
}
}
Add this to .cursor/mcp.json in your project root, or add globally in Cursor settings.
{
"mcpServers": {
"deeprecall": {
"command": "deeprecall-mcp",
"env": {
"DEEPRECALL_API_KEY": "ec_live_YOUR_KEY"
}
}
}
}
Add to your Windsurf MCP configuration file.
{
"mcpServers": {
"deeprecall": {
"command": "deeprecall-mcp",
"env": {
"DEEPRECALL_API_KEY": "ec_live_YOUR_KEY"
}
}
}
}
Add to Cline's MCP server settings in VS Code.
That's it. Start a new conversation. Your AI now has persistent memory.
Try saying things like:
The AI will automatically use Deep Recall tools to store and retrieve memories across sessions.
Pro tip: Drop this CLAUDE.md snippet into your project root to teach the AI to use memory proactively — searching at the start of every conversation and remembering important facts automatically.
Store facts, preferences, episodes, skills. Optionally tag with emotional context (happy, frustrated, excited).
Find memories by meaning, not just keywords. "outdoor hobbies" finds "loves hiking in mountains."
Store conversation turns. On paid plans, auto-extracts facts and entities from the conversation.
Track people, organizations, places — and the relationships between them.
Auto-detects when new memories conflict with old ones. "The API uses Postgres" vs "The API uses SQLite" — flagged.
Memories that get used grow stronger. Unused memories fade. Your memory self-optimizes.
Intelligent forgetting runs automatically. Important memories resist decay. Noise fades.
See total memories, entities, usage, and plan limits.
Check your plan, usage percentage, and upgrade options.
Search boosted by emotional similarity. Memories stored during frustration surface when frustrated again.
See how your memories cluster into communities. Knowledge graph analysis with Louvain detection.
Those files are loaded into context every time. They eat tokens and have no structure. Deep Recall stores memories in a real database with semantic search — your AI retrieves only what's relevant, not everything.
Stored in your own isolated database. Not shared with anyone. Not used to train models. You can export or delete everything at any time.
Every memory has a salience score (0-1). Memories you access frequently resist decay. Memories you never look at gradually fade. When you're near your storage limit, dead memories get pruned automatically. The result: your memory store stays clean without you doing anything.
When your AI retrieves memories and uses them in a good response, those memories get stronger (higher salience). When retrieved memories aren't useful, they get weaker. Over time, your memory store self-optimizes — useful knowledge rises, noise sinks.
Need more? See pricing — plans start at $19/mo for 100,000 memories.
Don't use MCP? The same features are available as a REST API. Full interactive docs at api.deeprecall.dev/docs.
import httpx
dr = httpx.Client(
base_url="https://api.deeprecall.dev",
headers={"X-API-Key": "ec_live_..."}
)
# Store
dr.post("/v1/memories", json={
"content": "Alice prefers dark mode",
"person": "alice",
"kind": "preference"
})
# Search
results = dr.get("/v1/memories/search",
params={"q": "user preferences"}).json()
Email aidan@deeprecall.dev