Mission · Why this matters
Track & limits — you read what's really running
Your agent searches and orders. What's still missing: the overview afterwards. Today you let it track its own orders, learn to read the platform's English status strings — and get to know the limits your agent works within. At the end, your course agent is complete.
What you'll be able to do
- Track orders from the chat — with
get_order_statusandlist_orders. - Read the platform's status strings and map them to the German badges.
- Tell the two kinds of limits apart: your Auftrags-Limits and the platform's rate limits.
The last two checkboxes
- Open the MCP Client Tool node and add the Tools
get_order_statusandlist_ordersunder Tools to Include → Selected. - That's your course agent's final loadout — five Tools: search, read ratings, prepare one order, query a status, list orders.
Core idea
Five Tools are enough — and the selection is deliberate: everything that could
route around your Freigabe (such as the catalog's task_* Tools) stays
off. With this loadout your agent can prepare everything and look up everything,
but decide nothing. Deciding remains your Tool.
Your agent tracks its orders
Ask in the chat: Show me my recent orders. Your agent calls
list_orders and returns the list — you should find the practice
order from lesson 3 in it. Then follow up: What's the status of "Test: Fensterreinigung"?
Now get_order_status answers.
And you'll notice something: the platform answers in English. On the technical wire, the stations of the agent path are called:
prepared— the order is prepared; on the web you know it as the "Vorbereitet" badge.approval_requested— the platform has requested your Freigabe; the Messenger message is on its way.pending_approval— the order is waiting for your decision (badge "Genehmigung ausstehend").rejected— you declined: the practice order's final stop (badge "Abgelehnt").
In the chat, your agent often translates these strings on its own — convenient, but memorize the originals: the execution log and technical responses always carry the English strings. If you can read both sides, you can track any order without guesswork.
Two kinds of limits
Your self-built agent works inside exactly the same guardrails as any other:
- Your Auftrags-Limits from the Agentowner course apply unchanged — they are tied to the Token, not to the tooling. Finished software or your n8n build: "at most 5 orders per day, at most €500 per order" is checked automatically on every
prepare_order. - The platform's rate limits: the MCP access allows 30 requests per minute, your Token 60 per minute in total. Plenty for a chat agent — a workflow in a loop can still tear through it. The demo key from the docs is capped tighter (10 per minute).
If it doesn't work: the two error signals
- 401 — ID missing: n8n reports "Authorization failed - please check your credentials", the platform says "Bearer token required". The Token is missing, wrong, or revoked — check the Bearer Auth credential or create a new Token for your agent.
- 429 — too fast: HTTP 429 (Too Many Requests) means: rate limit reached. Not a defect — your agent overdid the pace. Wait briefly and it continues; in looping workflows, build in a pause.
Your turn
Test yourself — retrieving from memory is what builds real retention.
Knowledge check
Your agent reports pending_approval for an order. What does that mean?
pending_approval means "Genehmigung ausstehend" (approval pending):
the platform has requested your Freigabe, and nothing happens until you decide. The
order is only approved after your yes — and nothing here has failed or been
dropped: this status means waiting, not failing.
Course complete — what now?
Your agent stands: five Tools, your Token, your Freigabe as the final authority. If you want to keep building, the platform's API docs are your reference — there you'll find all Tools in detail and the SDK, in case you ever want to develop directly without n8n.
