Who can use AfterShip Post-purchase MCP?
AfterShip Post-purchase MCP is useful for Customer Support, eCommerce Operations, and Logistics teams who regularly use AfterShip to check shipments or returns.
No coding or developer setup is required.
Overview
AfterShip Post-purchase MCP connects your AfterShip account with an AI assistant, such as Claude, letting you ask questions about your shipments and returns and get answers using your latest AfterShip data. You just need to connect once (a one-time sign-in).
For example, you can ask:
“Where is order #12345?”
“List all returns waiting for approval this week.”
The assistant retrieves the relevant information from AfterShip and presents it in everyday language, so you don't need to switch between dashboards or manually search for tracking or return details between tabs.
As of this release, the assistant can also take actions on your AfterShip data. For example, it can update shipment details, retrack a stalled shipment, mark a shipment status, approve or reject a return, record received return items, and resolve a return.
For actions that change your data, the assistant shows you what will change and waits for your explicit approval before making the change.
What is MCP?
MCP (Model Context Protocol) is an open, industry-standard adapter that allows compatible AI assistants to connect with business tools such as AfterShip.
You don't need to understand how MCP works to use it. Think of it like a USB port that lets compatible AI assistants connect to AfterShip and access the actions AfterShip makes available. Learn more at Model Context Protocol.
How AfterShip Post-purchase MCP works
You won't see most of this process happening, but here's what happens when you ask your AI assistant about your shipments or returns:
You ask the assistant a question or give it a task.
The assistant selects the appropriate AfterShip tool.
If the selected tool would change something, the assistant shows you what it is about to change and waits for your explicit approval.
AfterShip retrieves the information or performs the approved request against your live data and sends the result back.
The assistant uses the result to provide an answer in plain language.
What you can ask the assistant
AfterShip Post-purchase MCP provides 22 tools: 10 read tools and 12 write tools. Read tools run on their own; write tools require your confirmation before making changes.
Your assistant automatically chooses the appropriate tool based on your request.
Area | Type | What you can ask your AI assistant to do | Tools used |
Tracking | Read | Find and filter shipments by status, carrier, service, date, origin/destination, SKU, and more.
You can also open a shipment to view its full checkpoint timeline, sub-status, and estimated delivery date. |
|
Tracking | Write | Correct shipment details such as customer name, email, phone, note, tags, custom fields, shipment title, or language.
You can also retrack a stalled shipment or manually mark one as delivered, lost, or returned to sender. |
|
Returns | Read | Search RMAs by status, refund status, outcome, reason, value, or date. Open a return to view its items, reasons, exchange details, refund progress, receiving records, and history.
You can also read return workflow and allow/block list information. |
|
Returns | Write | Move a return through its lifecycle — approve or reject it, record received return items with inspection grades, and resolve it.
You can also add an internal comment and maintain the allowlist and blocklist. |
|
How the assistant handles changes
The write tools change real merchant and customer records. The assistant follows these rules when performing write operations:
Confirm first. Before any change, the assistant shows you exactly what will change — which record, which field, and from what value to what value — and waits for your explicit approval.
Never invent a value. If something is ambiguous, the assistant asks you to provide the value. Customer name, email, and phone changes must be confirmed verbatim. A rejection reason and inspection grade must come from you.
Batch operations show their scope. For anything affecting more than one record, the assistant lists the candidates and the scope of the change before acting.
The return lifecycle cannot skip steps. A return moves through the supported lifecycle: submitted > approved > received > resolved, or submitted > rejected. The assistant checks the current state before acting.
No promises beyond the tool result. The assistant reports the state returned by AfterShip. For example, it will not claim that a refund has been completed or an exchange has been arranged unless the tool confirms that state.
What cannot be undone through the connector
There is no tool to:
Un-approve, un-reject, or un-receive a return
Reopen a resolved return
Delete a return comment
Reverse a manually marked shipment status
receive_return_items is also not idempotent. Each call records an additional receiving batch, so calling it twice can double-count stock.
If the assistant is unsure whether a receive_return_items call went through, it will not automatically retry the operation. It re-reads the return instead.
Common use cases
Below are the everyday jobs the connector handles best, with example prompts you can copy.
1. Catch shipments that need attention
Value: Surface delivery problems — exceptions, failed delivery attempts, expired cases, and other issues — before customers escalate, so support and logistics teams can act first.
Prompt:
Show me all exception shipments from the last 7 days and group them by carrier.
Also try:
Find all expired shipments from the last 30 days.
2. Look up one order or customer
Value: Answer “where is my package?” in seconds by tracking number, order number, or customer email, with the latest status and full checkpoint timeline. Values must be complete: a partial tracking number or email fragment will not resolve.
Prompt:
Check the tracking details for order number #12345.
Also try:
Where is tracking number 123456789 now?
3. Search and summarize returns
Value: Get a fast operational view of your returns queue, grouped by status, outcome, or value, for daily review and follow-up.
Prompt:
List all returns pending approval from the last 7 days.
Also try:
Show me returns submitted this month and summarize them by status.
4. Review one return in full
Value: Understand a single return end-to-end — items and reasons, exchange details, refund progress, and key dates — without opening the admin.
Prompt:
Show me the details for RMA #RMA12345.
Also try:
Check whether order #12345 has a return request.
5. Fix a shipment record without opening the admin
Value: Correct a shipment detail in the same conversation where you found the issue, such as a typo in the customer's email, a wrong destination, or a tag for a follow-up queue.
Write — confirmed before it runs.
Prompt:
Tracking number 1Z999AA10123456784 has the wrong customer email — change it to [email protected].
Also try:
Add the tag priority-followup to the shipment for order #12345.
6. Process a return end to end
Value: Take a return through supported lifecycle steps without switching tools — approve it, record the parcel arriving with inspection grades, and close it out.
Write — confirmed before it runs. Refund outcomes only; exchanges are not supported yet.
Prompt:
Approve RMA #RMA12345.
Also try:
All three items on RMA12345 arrived — record two as grade A and one as grade D.
7. Explain why a return was handled that way
Value: Answer questions such as “Why did this RMA get approved automatically?” by reading your organization's own return automation rules alongside the return.
Prompt:
Why was RMA #RMA12345 approved automatically?
Also try:
Which return automation rules are running right now, and which are switched off?
8. Maintain the returns allow and block lists
Value: Act on return fraud or grant a VIP exemption in the same conversation where you found the pattern.
Write — confirmed per entry, and the reason must come from you.
Prompt:
Show me the current returns blocklist.
Also try:
Add [email protected] to the blocklist — reason: five high-value returns never shipped back.
Tip:
Make it automatic. In assistants that support scheduling, such as Claude, you can turn read-only checks into recurring tasks.
For example: “Every day at 9 AM, list any exception or failed-attempt shipments from the last 24 hours; say 'all clear' if there are none.”
Keep scheduled runs read-only because an unattended run has nobody available to confirm a write.
What you need before you start
Before connecting AfterShip Post-purchase MCP, make sure you have:
The Claude app, either Claude Desktop or Claude.ai in a browser, signed in.
An AfterShip account with access to the organization you want to connect.
About two minutes to complete the one-time secure sign-in using OAuth.
You don't need to enter a password or key in a settings file; just approve access through a browser window.
Connect AfterShip Post-purchase MCP to Claude
You can connect the connector from Claude's Connectors settings. Here are two ways to do it:
Method 1: Connect from the Connectors directory
This is a recommended option:
In Claude, go to Settings > Connectors.
Search for AfterShip Post-purchase.
On the results, click Connect.
If the connector doesn't appear in the search results yet, use the manual setup method below.
Method 2: Add the connector manually
In Claude, go to Settings > Connectors.
Click Add custom connector.
Enter the following details:
Name:
AfterShip Post-purchaseRemote MCP server URL:
https://mcp.aftership.com/post-purchaseOAuth Client ID:
claudeOAuth Client Secret: Leave empty.
Click Add.
Sign in to AfterShip
After adding the connector:
Click Connect.
In the browser window that opens, sign in to AfterShip.
Select the AfterShip organization you want the assistant to access.
Return to Claude.
Your connector is now connected.
Verify the connection
To verify, start a new conversation in Claude and ask:
What tools does the AfterShip Post-purchase connector provide?
The assistant should list 22 tools, including tools such as list_shipments, get_shipment, list_returns, get_return, approve_return, and update_shipment, without running any of them.
Important things to know
Reads run automatically; writes require your approval. 10 of the 22 tools read data and run on their own. The other 12 change real records, and the assistant shows you what will change and waits for your explicit approval first.
Your data is live. Answers come from the information currently available in AfterShip.
Scoped to one organization. The connector only accesses the AfterShip organization you selected during sign-in.
Scoped to your own role. The assistant can only perform actions allowed by your own AfterShip role. Connecting an assistant does not grant it more access than you already have.
Plan-gated. Tools require an active plan for the product they touch. If the organization has no Tracking or Returns plan, the relevant tools return a subscription link instead of data.
Streamable HTTP only. SSE and stdio transports are not supported. Set the transport type accordingly when configuring a client manually.
Some changes are permanent. Approving, rejecting, receiving, and resolving a return cannot be undone through the connector. A manually marked shipment status also cannot be reversed through the connector.
Not locked to one assistant. Because MCP is an open standard, the same connector can work with compatible AI assistants as they add support.
Troubleshooting connection issues
Issue 1: The "Add custom connector" option isn't available
Update Claude to the latest version.
Reopen Settings.
If you're using a Team or Enterprise plan, adding connectors may be limited to admins. Ask your organization's owner to add the connector to the workspace.
Issue 2: The sign-in window doesn't open
Firstly, check that your browser isn't blocking pop-ups from Claude.
Click Connect again.
Issue 3: The assistant shows data from the wrong organization
Go to Settings > click Connectors.
Remove the AfterShip Post-purchase connector and add it again.
During sign-in, select the correct AfterShip organization.
Issue 4: A tool returns a subscription link instead of data
The organization does not have an active plan for the product the tool is trying to access.
Tracking tools require a Tracking plan.
Returns tools require a Returns plan.
Issue 5: The assistant keeps asking for a full tracking number or email
Locators use exact matching. Partial tracking numbers or email fragments do not resolve.
Provide the complete tracking number or email address.
Issue 6: The assistant refuses to reject a return or record an inspection grade
The assistant cannot invent a rejection reason or infer an inspection grade. Provide the rejection reason or inspection grade explicitly.
If you're still unable to connect, contact AfterShip Support.
Key takeaways
AfterShip Post-purchase MCP lets you ask an AI assistant about your shipments and returns using your live AfterShip data.
The connector now provides 22 tools: 10 read tools and 12 write tools.
Read tools can retrieve information without additional approval. Write tools require your explicit approval before making changes.
You can connect AfterShip Post-purchase MCP to Claude through the Connectors directory or add it manually.
The connection is scoped to the AfterShip organization you select during sign-in and follows your existing AfterShip role permissions.
Once connected, you can find shipments, review tracking details, search and review returns, update shipment information, process supported return actions, and manage return allow/block lists.
For more technical information about AfterShip Post-purchase MCP, see the AfterShip Post-purchase MCP developer documentation.