Tasks

Create your first memory

A memory is the smallest unit of knowledge in knowmind. It can be a note, a meeting memo, a definition or a list. This guide shows three ways: in the dashboard, from the command line, via the API.

Audience

All new users. No prior knowledge required. For the second and third path you additionally need the knowmind CLI or an API token.

Prerequisites

Active knowmind account (see Quickstart)
Active workspace (created automatically during onboarding)
For path 2: knowmind CLI installed and signed in
For path 3: API token (available on every plan, including Private)

Steps

Path 1: In the dashboard

In the dashboard, in the Quick access area, click "Create memory" — or go directly to /dashboard/memory/new.

In the form, the following fields are required or recommended:

Field	Required	Note
Title	recommended	One line, no full stop
Content	yes	Markdown is supported
Type	recommended	semantic, episodic, procedural, reference
Tags	optional	Comma-separated, for filtering later
Source	optional	Free text, e.g. "Team meeting 2026-03-12"

Click Save.

Outcome: You land on the detail view of the new memory. In the dashboard, the statistics tile increments by one.

2
Path 2: From the command line
bash
```
# Short note straight from stdin
echo "The next quarterly report is due 30 June." \
  | knowmind upload - --title "Quarterly report 2026 Q2"

# Upload a file
knowmind upload ./meeting-notes.md --title "Workshop Müller 2026-03-12"
```
For longer text knowmind chunks the content into sensible sections, generates an embedding per section and creates a node in the Knowledge Graph.
Outcome: On success the command prints the memory ID and the number of sections created.

Path 3: Via the API

For your own applications, scripts or workflow tools such as n8n. Writing goes through the MCP endpoint POST /api/mcp/v1 (tools/call knowmind_store_memory) — the same persistent path used by the CLI and AI clients. The entry is indexed durably (Postgres + knowledge graph).

bash

curl -X POST https://knowmind.de/api/mcp/v1 \
  -H "Authorization: Bearer kmt_…" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "knowmind_store_memory",
      "arguments": {
        "title": "QM training mandatory",
        "content": "All service staff complete the QM training annually.",
        "memory_type": "semantic",
        "tags": ["qm", "training"],
        "source": "internal-handbook"
      }
    }
  }'

Outcome: A JSON-RPC response with the memory_id of the durably indexed entry. A near-identical entry is detected and not stored twice (deduplicated: true referencing the existing ID).

Verify the result

The memory is in place when all three conditions hold:

The dashboard "Recent activity" lists "Memory created".
The Memories statistics tile incremented by one.
A search with a word from the content returns the new memory.

Troubleshooting

Error message	Cause	Resolution
Save button stays disabled (dashboard)	Content field is empty, or the plan limit is reached.	Enter content. If the cap is reached (2,500 memories on Private), upgrade under Plan.
knowmind upload: 403	Token scope does not include write.	Edit the token in the dashboard and add scope write.
API responds 422 invalid memory_type	Memory type outside the allowed set.	Allowed values: `semantic`, `episodic`, `procedural`, `reference`, `working`, `entity`.
Memory shows up, but search does not find it	Indexing takes a few seconds. For very short notes the semantic hit may be absent — the full-text BM25 index then carries the result.	Search for a literal word from the content. If consistently missing, check the memory count with `knowmind stats` and service availability with `knowmind health`.