your memory layer is live.

supermemory is running on this machine. embeddings, storage, and search all happen locally — nothing leaves your laptop.

endpoint

http://localhost:6767

api key

sm_przH5…BQEx

data directory

/opt/supermemory/data

build

Build with supermemory local

Use the same SDK and REST API as the cloud, just pointed at your machine. Whatever you build here is portable — flip the base URL and ship it to production.

install the sdk

npm install supermemory
# or: bun add supermemory · pnpm add supermemory

use it

import Supermemory from "supermemory"

const client = new Supermemory({
  apiKey: "sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx",
  baseURL: "http://localhost:6767",
})

await client.memories.add({
  content: "running on my laptop. zero cloud.",
})

const results = await client.search.execute({
  q: "what's running on my laptop?",
})

curl, if you prefer

curl -X POST http://localhost:6767/v3/documents \
  -H "Authorization: Bearer sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx" \
  -H "Content-Type: application/json" \
  -d '{"content": "supermemory is running locally!"}'

use

Use with your agents

Point your existing tools at this local instance. Click any plugin for setup steps — base URL and key are pre-filled for you.

Claude Code Persistent memory across coding sessions. Auto-recalls relevant context at session start.

Point Claude Code at this server

Add to your shell profile so the plugin uses your local instance:

export SUPERMEMORY_BASE_URL="http://localhost:6767"
export SUPERMEMORY_CC_API_KEY="sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx"

Install the plugin

Inside a Claude Code session, run:

/plugin marketplace add supermemoryai/claude-supermemory
/plugin install claude-supermemory

OpenCode Semantic search across previous sessions, auto-capture of coding decisions.

Point OpenCode at this server

Add to your shell profile:

export SUPERMEMORY_BASE_URL="http://localhost:6767"
export SUPERMEMORY_API_KEY="sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx"

Install the plugin

Use --no-tui for non-interactive environments:
```
bunx opencode-supermemory@latest install
```
Verify ~/.config/opencode/opencode.jsonc

Should include:
```
{
  "plugin": ["opencode-supermemory"]
}
```

OpenClaw Multi-platform memory across Telegram, WhatsApp, Discord, Slack.

Install the plugin

From your OpenClaw project:

openclaw plugins install @supermemory/openclaw-supermemory

Configure for local

Run setup, paste the api key shown above, and enter the local base URL when prompted:

openclaw supermemory setup
# base url:  http://localhost:6767
# api key:   sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx

Hermes Memory for the Hermes agent. Semantic search and conversation capture.

Start the memory wizard

On the host running Hermes:
```
hermes memory setup
```

Choose Supermemory, then point it at local

When prompted, paste the base URL and api key shown above:

base url: http://localhost:6767
api key:  sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx

Cursor Persistent AI memory for Cursor via MCP.

Install the MCP server

Cursor will pick up the local MCP automatically. Add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "supermemory-local": {
      "url": "http://localhost:6767/mcp",
      "headers": { "Authorization": "Bearer sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx" }
    }
  }
}

OpenAI Codex Persistent memory for the Codex CLI. Recalls context before each prompt.

Point Codex at this server

Add to your shell profile:

export SUPERMEMORY_BASE_URL="http://localhost:6767"
export SUPERMEMORY_CODEX_API_KEY="sm_przH5DVWq3KXd21dzCBWnN_sxtl71hqdSSurAZgWnjDez8tM1A2Xiwy9NrtaSMnu3okysLZLQWkv8y0tygMBQEx"

Install the plugin

Run:
```
npx codex-supermemory@latest install
```

openapi spec api reference docs ↗ github ↗