HighCohesion MCP Server prompt library
Copy-ready prompts for finding records, understanding failures, checking data and preparing requests, grouped by task.
Ready-to-use prompts for the questions HighCohesion customers ask most often, grouped by what you are trying to do. Copy a prompt into your AI assistant, replace everything in [square brackets], and send it. For tips on getting better answers, see How to get good answers from the HighCohesion MCP Server.
Every prompt here only reads data. None of them replicate, change or delete anything.
Find the right prompt
Find a record
When an order, PO, transfer or product has not arrived where you expected.
Where is this record?
Use it when: An order, PO, transfer or product has not reached the system you expected.
You will need: The reference exactly as it appears in the source system, including any prefix such as #.
Find where [order, PO, transfer or SKU reference] has got to. I expected it to reach [system]. For each stream it passed through, show the status and time. If it stopped, tell me where and why: failed, still pending, filtered out by the stream's rules, or never picked up. Give me the event ID for anything that failed and tell me what to do next.A good answer includes: Each stream the record passed through, the point it stopped, and a reason backed by an event or a stream rule.
Why was this not picked up?
Use it when: There is no event at all for a record, so it never entered HighCohesion.
You will need: The reference, and what you know about the record: its status, tags, payment method, location or anything unusual about it.
There is no event in HighCohesion for [reference], which should have gone from [source system] to [destination system]. Here is what I know about the record: [status, tags, payment method, location, anything unusual]. Check the rules the stream uses to pick up records (filters, statuses, tags, dates, locations, order types) and tell me which rule it most likely failed. Then tell me what would need to change for it to be picked up.A good answer includes: The specific rule the record did not meet, and the change that would let it through.
Is it safe to replicate this?
Use it when: Before you replicate a failed event from the Control Panel.
You will need: The reference and the stream name.
I want to replicate the failed event for [reference] on [stream name]. Before I do, check: has the same record already succeeded on this stream, so replicating would create a duplicate? Has the cause of the failure been fixed, or will it fail again? Is there a newer event for the same record? Give me a clear go or no-go with the reason.A good answer includes: A go or no-go, with the evidence for each check.
Understand a failure
When you have an error message or an alert and need to know what it means and who acts.
Explain this error
Use it when: An event has failed and the error message does not make sense.
You will need: The error text, or the event or issue ID from the Control Panel or the alert email.
Explain this error in plain English: [paste the error message, or give the event or issue ID]. Tell me what happened, which system rejected it and the most likely cause, quoting the part of the data HighCohesion sent that caused it. Tell me who needs to fix it: us in our own system, HighCohesion, or the other system. Then tell me whether other records have hit the same error in the last 7 days.A good answer includes: A plain cause, the rejecting system, who owns the fix, and whether it is a one-off.
Note: Reading the data inside events is coming in the next few weeks. Until then, the assistant can explain the error message but cannot quote the data that caused it.
Do I need to act on this alert?
Use it when: An alert email has arrived and you are not sure whether it needs you.
You will need: The alert email, pasted in full.
I received this HighCohesion alert: [paste the alert email]. Tell me in two sentences what it means, whether the problem is still happening or has cleared, and whether I need to do anything. If the record has been cancelled or changed since, tell me whether the alert will keep repeating.A good answer includes: A yes or no on action, and whether the alert will stop on its own.
Why is this stream degraded?
Use it when: A stream shows as degraded in the Control Panel.
You will need: The stream name.
The stream [stream name] shows as degraded. Explain what that means for this stream right now, which errors are behind it, when they started, and whether records are still getting through.A good answer includes: The errors driving the status, when they started, and whether data is still flowing.
Deal with a lot of failures
When many records fail at once or alerts pile up.
Summarise failures
Use it when: You want to know whether many failures are one problem or several.
You will need: A stream name (or none, for all streams) and a time window.
Summarise failed events on [stream name, or all streams] over [time window, e.g. the last 24 hours]. Group them by underlying cause, not exact error text. For each group give the cause in plain English, how many failed, which streams, first and last seen, and three example references. Say whether each group looks like one root cause, such as expired credentials or a system outage, or separate record problems. Order by impact and tell me which to deal with first.A good answer includes: A short list of causes, ranked, with counts and examples.
Which records does this issue affect?
Use it when: You know the problem and need the full list of records it hit.
You will need: The issue ID, or a description of the error, and a time window.
For [issue ID, or describe the error], list every record affected in [time window], with its reference, stream, first failure time and whether it has since succeeded. Show it as a table I can export, and tell me how many are still failing.A good answer includes: An exportable table and a count of records still failing.
Why am I getting so many alerts?
Use it when: Your inbox has filled with alerts in a short time.
You will need: Roughly how many alerts, since when, and the stream if they share one.
We have received [number] alerts from [stream name, or leave blank] since [time]. Tell me whether they come from one problem or several, whether they are repeat alerts for the same records, what triggered the spike, and whether it is still happening.A good answer includes: Whether it is one problem, whether it is still going on, and what set it off.
Check the data
When a record went through but something about it looks wrong.
Compare what came in with what went out
Use it when: A quantity, price, tax amount, batch or status looks wrong in the destination.
You will need: The reference, and the field you are worried about if you know it.
For [reference], compare what HighCohesion received from the source with what it sent to the destination. Focus on [field, or leave blank for quantities, prices, tax, currency, SKUs, batches and statuses]. Show a table of field, value received, value sent, and whether it changed. For each difference, explain which mapping or lookup table caused it, and flag anything that looks wrong rather than intended.A good answer includes: A field-by-field table and the mapping behind each change.
Note: Reading the data inside events is coming in the next few weeks. Until then, the assistant can confirm each event's status but cannot compare field values.
Explain a stock change
Use it when: Stock for a SKU moved and you do not know why.
You will need: The SKU, the system and location, the size of the change and roughly when.
Stock for [SKU] in [system and location] changed by [amount] around [date and time]. Find the events that changed it, show each with time, stream, quantity sent and whether it was an adjustment or a stock level, and tell me which one caused the change.A good answer includes: The event that caused the change, with the quantity it sent.
Note: Reading the data inside events is coming in the next few weeks. Until then, the assistant can find the events that changed stock but may not be able to show the quantities they sent.
Why did this record do that?
Use it when: A record went through but behaved unexpectedly: a cancelled order synced, a shipment split, a duplicate appeared.
You will need: The reference, what happened, and what you expected.
[reference] went through, but [what happened] when I expected [what you expected]. Walk through what HighCohesion received and sent for this record at each step, and tell me which rule in the stream produced this result. Say whether it is working as configured or looks like a fault.A good answer includes: The rule responsible and a clear view on whether it is by design.
Find other records with the same problem
Use it when: You have found one bad record and want to know how widespread it is.
You will need: The reference, the problem, the stream and a time window.
[reference] has this problem: [describe it, e.g. missing tax line, zero quantity]. Check other records on [stream name] over [time window] and list any that show the same pattern, so we can fix them together.A good answer includes: A list of matching records, or confirmation that it is a one-off.
Understand your setup
When you need to know how an integration works before you change or rely on it.
How does this stream work?
Use it when: You need to understand what a stream picks up, what it sends and where.
You will need: The stream name.
Explain how the stream [stream name] works, for someone who does not write code. Cover what triggers it and which records it picks up, what it sends and where, the key field mappings with field names as they appear in each system, any lookup tables, and what happens when an event fails. Then answer: [your question].A good answer includes: Triggers, filters, mappings and failure behaviour in plain English.
Where does this field come from?
Use it when: A value in the destination system is not what you expected and you want to know where it is set.
You will need: The stream name and the field name as it appears in the destination.
On [stream name], where does [field name in the destination] come from? Show the source field, any mapping or lookup table applied, and a worked example from a recent record.A good answer includes: The source field, the logic in between, and a real example.
Note: Reading the data inside events is coming in the next few weeks. Until then, the assistant can show the mapping but not a worked example from a real record.
Which streams involve this system or data?
Use it when: You are planning a change, such as a new warehouse or WMS, and need to know what it touches.
You will need: The system name or the type of data.
List all our streams that [involve a system, or carry a type of data such as inventory], with source, destination, status and a one-line description of each.A good answer includes: A complete list you can plan from.
Compare two streams
Use it when: Two streams should behave the same way, for example for two stores or countries, but do not.
You will need: Both stream names.
Compare [first stream] with [second stream]. Show where their filters, mappings and lookup tables differ, as a table, and flag differences that look unintentional.A good answer includes: A table of differences with the likely unintended ones flagged.
Look up reference data
Use it when: You need to check or export a lookup table, such as shipping methods or locations.
You will need: The lookup table name, and a value if you want one looked up.
Show the contents of the [lookup table name] lookup table as a table I can export. Then tell me what [value] maps to.A good answer includes: The table, and the mapped value.
Prepare a request for us
When you need HighCohesion to act. A complete request gets resolved faster.
Draft a change request
Use it when: You want a stream to behave differently.
You will need: The change in your own words, and an example record if you have one.
Help me write a change request for HighCohesion support. I want: [the change, in your own words]. Example record: [reference]. Find the exact names and IDs of the streams affected, describe how they behave now, and write the request with a subject line, affected streams, current behaviour, requested behaviour, example record, business impact and deadline. Ask me for the impact and deadline if I have not given them. Flag any other streams likely to need the same change.A good answer includes: A request our team can act on without coming back to you.
Write up a problem for support
Use it when: Something is wrong and you need our help to fix it.
You will need: A description of the problem and any references you have.
Write a support request for HighCohesion about this problem: [describe it]. Include the stream name and ID, example references with event IDs, the error, when it started, how many records are affected, what I have already checked, and the business impact. Ask me for anything you cannot find.A good answer includes: Stream, examples, error, scale and impact in one message.
Routine checks
For regular check-ins, so you spot problems before your customers do.
Morning check
Use it when: At the start of the day.
You will need: Nothing, or a start time.
Give me a morning check of our integrations: any streams paused, degraded or not running as expected, failures since [e.g. 6pm yesterday], and anything still failing after retries. Keep it to what needs action today.A good answer includes: A short list of what needs action today, or confirmation that nothing does.
Weekly summary for my team
Use it when: At the end of the week, for colleagues who do not use the Control Panel.
You will need: The audience and the date range.
Write a short weekly summary of our integrations for [e.g. the ops team] covering [date range]: failures by stream and cause, anything still open, and anything that changed compared with the week before. Plain English, no jargon.A good answer includes: A summary you can forward as it is.