Every way to plug AgentRAM into your agent: a plain REST API, official Python and TypeScript SDKs, and an MCP server for clients like Claude Desktop. Pick your stack, copy one line, and your agent has persistent memory.
agentram_...) is emailed to you.The official Python client. Zero third-party dependencies, standard library only.
Installs as agentram-sdk, imports as agentram.
from agentram import AgentRAM
ram = AgentRAM(api_key="agentram_...", agent_id="agent-01")
ram.store("user_language", "French")
ram.recall("user_language") # "French"
The official TS/JS client. Native fetch, full type declarations, and it runs anywhere: Node 18+, Deno, Bun, browsers, and edge runtimes.
import { AgentRAM } from "agentram-sdk";
const ram = new AgentRAM({ apiKey: "agentram_...", agentId: "agent-01" });
await ram.store("tone", "formal");
await ram.recall("tone"); // "formal"
Give MCP-compatible clients — Claude Desktop, Cline, Cursor, and others — a set of memory tools that map one-to-one to the API. Ten tools in total: store, recall, list, search, and delete, plus check-credits and the shared-namespace equivalents.
Add this to your client's MCP config, then restart it. For Claude Desktop on macOS the file is ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"agentram": {
"command": "npx",
"args": ["-y", "agentram-mcp"],
"env": {
"AGENTRAM_API_KEY": "agentram_..."
}
}
}
}
No SDK required. Anything that can make an HTTP request works. Send your key in the x-api-key header and you are done.
curl -X POST https://api.agentram.dev/memory \
-H "x-api-key: agentram_..." \
-H "Content-Type: application/json" \
-d '{"agent_id":"agent-01","key":"user_pref","value":"prefers short answers"}'
Using a framework? These walk through AgentRAM inside a real setup, built on the SDK and REST API above.
© 2026 AgentRAM. All rights reserved.