memory_store_session
Store a full multi-turn conversation as one memory unit so agents can preserve context and retrieve entire sessions reliably instead of isolated turns.
Instructions
Store a full conversation session as a single memory unit.
Use this instead of memory_store when you want to preserve the full context of a multi-turn conversation. All turns are stored together, making session-level retrieval more reliable than storing individual turns separately.
Example: { "turns": [ {"role": "user", "content": "How do I configure Redis?"}, {"role": "assistant", "content": "Set REDIS_URL in your .env file..."} ], "session_id": "optional-stable-id", "tags": "redis,configuration" }
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"turns": "<turns>"
},
"name": "memory_store_session"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| tags | No | Additional tags (comma-separated string or array). 'session:<id>' is always added automatically. | |
| store | No | Target store partition (default: 'default'). Use 'docs' for documents, 'all' for cross-store search. | |
| turns | Yes | Ordered list of conversation turns. | |
| agent_id | No | Optional authoring agent id. Falls back to the X-Agent-ID header, then MCP_AGENT_ID. Passing it here makes the caller's author win over the server default. | |
| metadata | No | Optional extra metadata. | |
| session_id | No | Optional stable identifier for this session. Auto-generated UUID if omitted. |