01What it is
A remote MCP memory endpoint for AI agents. Connect over hosted MCP and your agent can store and recall memory without you building storage, recall, stale-fact handling, audit receipts, deletion logic, or permissions.
02Why it matters
Agents need memory that is governed, scoped, auditable, and deletable — not just a vector store. AgentPrizm ships confidence scores, fact-validity windows, audit receipts, container scopes, and right-to-forget on every plan.
03Supported clients
Claude Code, Cursor, Claude Desktop, and any MCP-capable agent. Paste one config block:
{
"mcpServers": {
"agentprizm-memory": {
"type": "http",
"url": "https://agentprizm.com/api/mcp",
"headers": { "Authorization": "Bearer ap_..." }
}
}
}04Scopes and permissions
Memory is scoped per user, org, agent, and session; cross-scope queries are explicit and logged. Every recall returns an audit receipt, and any memory can be removed with one right-to-forget call.