How to Give Your AI Agent Persistent Memory in 5 Minutes

AgentBrain2026-08-17

Most AI agents forget everything the moment a session ends. Giving an agent persistent memory used to require building custom storage, embedding pipelines, and retrieval logic. With a hosted shared brain like AgentBrain, you can give any MCP-compatible agent durable, shared memory in about five minutes — no code changes.

What you'll get

Step 1 — Get a free API key

Register to get an API key on the free tier:

curl -X POST https://agentbrain-auth.onrender.com/auth/register \
  -H "Content-Type: application/json" \
  -d '{"email": "you@example.com", "password": "yourpassword"}'

You'll get back an API key that looks like ab_.... Keep it handy.

Step 2 — Connect your agent via MCP

Add AgentBrain as an MCP server in your agent's client config (Claude Desktop, Claude Code, Cursor, Windsurf, or any MCP client):

{
  "mcpServers": {
    "agentbrain": {
      "url": "https://agentbrain-auth.onrender.com",
      "headers": { "X-API-Key": "ab_your_api_key_here" }
    }
  }
}

Step 3 — Store a memory

Once connected, your agent can call the remember tool. You can also test it directly:

curl -X POST https://agentbrain-auth.onrender.com/mcp/remember \
  -H "Content-Type: application/json" -H "X-API-Key: ab_your_api_key_here" \
  -d '{"content": "The user prefers concise answers and works in fintech.", "metadata": {"tag": "preference"}}'

Step 4 — Recall it later

In any future session, the agent retrieves relevant memories with recall:

curl -X POST https://agentbrain-auth.onrender.com/mcp/recall \
  -H "Content-Type: application/json" -H "X-API-Key: ab_your_api_key_here" \
  -d '{"query": "what does the user prefer?"}'

Step 5 — Verify your limits

Check your tier and remaining capacity:

curl -H "X-API-Key: ab_your_api_key_here" \
  https://agentbrain-auth.onrender.com/auth/limits

Going further

The free tier is enough to get started. For higher memory limits, knowledge-graph tools, and agent discovery, see the paid plans. To share context across a whole team of agents, register each agent with register_agent and let them discover each other with discover_agents.

Key takeaways