> ## Documentation Index
> Fetch the complete documentation index at: https://docs.memanto.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# memanto remember

> Store a new memory for the active agent — single, batch, or extracted from a conversation.

# memanto remember

Store a new memory for the active agent. Supports three input modes: a single memory, a batch of memories from a JSON file, or memories auto-extracted from a conversation transcript.

```bash theme={null}
memanto remember TEXT [OPTIONS]
```

**Arguments:**

* `TEXT` - Memory content (required unless `--batch` or `--from-conversation` is used)

**Options:**

* `-t, --type TEXT` - Memory type (fact, preference, goal, decision, artifact, learning, event, instruction, relationship, context, observation, commitment, error)
* `--title TEXT` - Memory title (defaults to truncated content)
* `-c, --confidence FLOAT` - Confidence score 0.0-1.0 (default: 0.8)
* `--tags TEXT` - Comma-separated tags
* `-s, --source TEXT` - Source of the memory, e.g. `user`, `agent_name` (default: `user`)
* `-p, --provenance TEXT` - Provenance/origin of the memory, e.g. `inferred`, `corrected` (default: `explicit_statement`)
* `--batch PATH` - Path to a JSON file with an array of memory objects (batch mode)
* `--from-conversation PATH` - Path to a JSON conversation file, or `-` to read from stdin (conversation-extraction mode)
* `--dry-run` - Preview extracted conversation memories without storing them (only valid with `--from-conversation`)
* `--max-memories INTEGER` - Maximum memories to extract from a conversation (default: 20, 1-100)
* `--ai-model TEXT` - Optional model override for conversation extraction

**Examples:**

Simple memory:

```bash theme={null}
memanto remember "Paris is the capital of France"
```

With type:

```bash theme={null}
memanto remember "Customer prefers email" --type preference
```

With title:

```bash theme={null}
memanto remember "Will deliver report Friday" --type commitment --title "Delivery Deadline"
```

Lower confidence:

```bash theme={null}
memanto remember "Thinks customer is in finance" --type fact --confidence 0.7
```

With tags, source, and provenance:

```bash theme={null}
memanto remember "Customer moved to marketing team" \
  --type fact \
  --tags "vip,enterprise" \
  --source agent \
  --provenance inferred
```

Batch from file:

```bash theme={null}
memanto remember --batch batch.json
```

Extract from a conversation file:

```bash theme={null}
memanto remember --from-conversation chat.json
```

Extract from stdin:

```bash theme={null}
cat chat.json | memanto remember --from-conversation -
```

Preview extraction from conversation without storing:

```bash theme={null}
memanto remember --from-conversation chat.json --dry-run
```

Limit extracted memories and pin a model:

```bash theme={null}
memanto remember --from-conversation chat.json --max-memories 10 --ai-model anthropic.claude-sonnet-4-6
```

**Memory Types:**
`fact`, `preference`, `decision`, `commitment`, `goal`, `event`, `instruction`, `relationship`, `context`, `learning`, `observation`, `error`, `artifact`

**Notes:**

* `--from-conversation` cannot be combined with `TEXT` or `--batch` — use one input mode at a time.
* Conversation JSON must be an array of `{role, content}` message objects (up to 200 messages).
* Batch JSON must be an array of memory objects (each needs at least `content`), up to 100 items per file.
* Backed by the [Remember](/api-reference/data/remember), [Batch Remember](/api-reference/data/batch-remember), and [Extract Memories](/api-reference/data/extract-memories) API endpoints.