long-term memory
for AI agents.

import { Recall } from "@recall-ai/sdk";

const recall = new Recall({ apiKey: process.env.RECALL_API_KEY!, baseUrl: "https://<your-host>" });

await recall.save("user_42", "I moved to Toronto last month and I prefer Python.");
const { results } = await recall.find("user_42", "where does the user live?");

# save
curl -X POST https://<your-host>/v1/realms/user_42/file \
  -H "Authorization: Bearer $RECALL_API_KEY" -H "Content-Type: application/json" \
  -d '{"content": "I moved to Toronto last month and I prefer Python."}'

# recall
curl -X POST https://<your-host>/v1/realms/user_42/find \
  -H "Authorization: Bearer $RECALL_API_KEY" -H "Content-Type: application/json" \
  -d '{"q": "where does the user live?", "limit": 5}'

import os, httpx

recall = httpx.Client(base_url="https://<your-host>",
                      headers={"Authorization": f"Bearer {os.environ['RECALL_API_KEY']}"})
recall.post("/v1/realms/user_42/file",
            json={"content": "I moved to Toronto last month and I prefer Python."})
hits = recall.post("/v1/realms/user_42/find", json={"q": "where does the user live?"}).json()

npm install -g @recall-ai/cli

recall login                # opens your browser
recall status               # account, realm and usage
recall connect my-project   # link this folder to a realm

recall save "We deploy with Vercel."
recall find "how do we deploy?"
recall about                # core and changing facts
recall remove "old notes"   # asks before removing
recall files                # or: recall files <id>
recall files rm <id>        # delete a file
recall sync                 # run Sync now
recall realms               # all your realms

recall setup claude         # or cursor, codex, opencode
recall mcp                  # MCP server over stdio
recall explore              # the realm as files
recall explore RECALL.md    # print one file

# 1. sign in once (opens your browser)
npx @recall-ai/cli login

# 2. in Claude Code: recalls the project at session start and with each prompt,
#    keeps learning as the conversation grows, adds the Recall tools
/plugin marketplace add Ravicodess/Recall-AI
/plugin install recall@recall-ai

# 1. sign in once (opens your browser)
npx @recall-ai/cli login

# 2. add the Recall tools to ~/.cursor/mcp.json
npx @recall-ai/cli setup cursor

# 3. the rule that tells the agent when to find and save
mkdir -p .cursor/rules && curl -fsSL https://raw.githubusercontent.com/Ravicodess/Recall-AI/main/plugins/cursor/recall.mdc -o .cursor/rules/recall.mdc

# 1. sign in once (opens your browser)
npx @recall-ai/cli login

# 2. add the Recall tools to ~/.codex/config.toml
npx @recall-ai/cli setup codex

# 3. the rule that tells the agent when to find and save
curl -fsSL https://raw.githubusercontent.com/Ravicodess/Recall-AI/main/plugins/codex/AGENTS.md >> AGENTS.md

# 1. sign in once (opens your browser)
npx @recall-ai/cli login

# 2. add the Recall tools to ~/.config/opencode/opencode.json
npx @recall-ai/cli setup opencode

# 3. bring in and save notes automatically in this project
mkdir -p .opencode/plugin && curl -fsSL https://raw.githubusercontent.com/Ravicodess/Recall-AI/main/plugins/opencode/index.js -o .opencode/plugin/recall.js

# signed in with `npx @recall-ai/cli login`: the CLI runs the MCP server for you
{
  "mcpServers": {
    "recall": { "command": "npx", "args": ["-y", "@recall-ai/cli", "mcp"] }
  }
}

# or over HTTP, with a key from API keys
{
  "mcpServers": {
    "recall": { "type": "http", "url": "https://<your-host>/mcp", "headers": { "Authorization": "Bearer rc_your_key" } }
  }
}

# save pages and selections, find from the toolbar
1. Download plugins/browser from https://github.com/Ravicodess/Recall-AI/tree/main/plugins
2. chrome://extensions → Developer mode → Load unpacked
3. Open the extension settings and paste a key from API keys

Save a page: toolbar button, right-click, or Alt+Shift+S