What is Portmate MCP

Portmate MCP connects your AI agent to Portmate's logistics tools through the Model Context Protocol. It gives an authorized agent a structured way to request drayage quotes, create bookings from quotes, and retrieve shipment tracking information while you work in the agent's own interface.

You describe the task. Your agent works out which available tools to use, gathers the required information, and explains the results. Portmate validates the business request and performs the supported operation.

Portmate MCP is available in preview to customers with access upon request. To discuss access, contact the Portmate team with the agent you want to connect and the workflow you want to try.

For an importer, forwarder, or operations team, the practical value is continuity. A lane request can begin in the assistant you already use, become a quote you can review, and continue into booking and shipment questions without copying every result between separate conversations.

This guide explains what that connection does, how to use it thoughtfully, and where agent-connected logistics could go next. If you are new to the underlying service, start with what drayage means.

What MCP adds to an AI agent

The Model Context Protocol, or MCP, is an open standard for connecting AI applications to external systems. The official MCP introduction describes connections to tools, data sources, and workflows.

That connection matters because a conversational answer and a business operation have different requirements. An assistant may explain drayage from general knowledge. Obtaining a Portmate quote requires the supported pricing operation. Reporting a shipment's latest recorded milestone requires a shipment lookup.

MCP gives the application hosting your agent a common way to discover and call those operations. Each exposed tool has a description and an input schema that tells the agent what information the operation accepts. Results return information the agent can use in its next response. The MCP server concepts guide explains these building blocks.

Your agent, the connection, and Portmate

Three roles help make the arrangement clear:

• Your agent interprets your request, asks useful follow-up questions, selects tools, and explains what happened.
• The MCP connection carries the available tool definitions, structured requests, and returned results between the agent's application and Portmate.
• Portmate's services apply business rules, check required inputs and authorization, and perform the supported operation.

The application hosting the agent is often called an MCP host. It uses an MCP client to communicate with a server. You usually encounter that through a connector, plugin, integration setting, or configuration supplied by your agent's application. The official architecture overview describes those roles in more detail.

Your agent can therefore keep its own conversational style and working context while using Portmate for Portmate-specific facts and actions. The language model does not become the pricing engine, and a sentence saying that a booking succeeded does not establish success without the corresponding service result.

A request becomes a sequence of tool calls

A useful interaction generally follows this sequence:

1. You state an outcome, such as obtaining a quote for a specified container move.
2. The agent identifies the relevant tools and checks the information they require.
3. The agent resolves missing or ambiguous details, including which terminal or shipment you mean.
4. Portmate validates and executes the request within the connected customer's access.
5. The agent explains the returned outcome, including any assumptions, missing information, or next step.

One message can require several calls. For example, identifying a terminal and creating a quote are separate operations. A later request to book that quote is another operation with its own inputs and consequences.

What you can do through Portmate MCP

Quotes, bookings, and tracking are the central workflows. Portmate's tool layer also includes supporting operations for location research, accessorial rates, lane weight guidance, quote retrieval, article knowledge, and human review. Use the tools made available with your preview connection as the current guide to what your agent can call.

Request and review drayage quotes. Your agent can assemble the details for a supported U.S. container drayage move and request pricing. Useful information includes the loaded interchange location, facility location, import or export direction, container type, commodity, net cargo weight, and any requested services.

A clear starting request is:

Get a quote for one 40HC import container from APM Elizabeth to Linden, NJ 07036. The cargo is dry, nonhazardous packaged goods with a net weight of 18,000 kg. Explain the price assumptions and accessorial conditions. Get a quote only.

The result needs more than a total. Review the priced locations, equipment, cargo basis, included services, and applicable terms. If an assumption was used, the agent should identify it. For a fuller explanation of quote boundaries, read what an all-in drayage quote means and how to get one.

Retrieve quotes and request copies. An existing quote reference lets the agent retrieve the relevant quote rather than reconstruct its price from a previous conversation. When the tool is available and you request it, the agent can also arrange a quote copy. Looking up a quote, generating a new quote, and sending a copy are distinct requests; tell the agent which outcome you need.

Create a booking from a quote. After you select a quote, the booking workflow needs the required facility and container information. The agent can collect the missing details and submit the supported booking operation when you authorize it. The returned shipment reference gives you something concrete to retain and use for later questions. The booking guide explains the underlying quote-to-booking process.

Ask shipment tracking questions. A Portmate shipment reference, quote number, or container number can identify a shipment for a tracking lookup. Ask about recorded progress, the next stop, and an available next-stop ETA. A useful prompt is Check the latest tracking information for this shipment. Separate completed milestones from planned events and estimates.

The answer depends on the available record. A container number alone does not establish that Portmate has a matching shipment. For more ways to ask operational questions, see how to track your shipment in chat.

Research locations and operating requirements. Supporting tools can help identify terminal or depot candidates, find nearby supported metros, retrieve accessorial rates, and evaluate lane weight support. These checks answer different questions. A nearby location is not proof of equipment availability; a published accessorial rate is not proof that a charge occurred; a weight advisory is not a permit or final route approval.

Use the closest-terminal guide for location research and the lane weight guide for the information needed to assess cargo weight support.

Get explanations or request human review. Article retrieval can help an agent explain terminology and product workflows. A supported human-review tool can submit a defined shipment request when that is the appropriate next step. Finding an article about a service does not establish that the service can be quoted or booked automatically.

The instant quote operation focuses on supported U.S. container drayage between a loaded port or rail interchange and a shipper or consignee facility, with the relevant empty-equipment movement. Other transport or special-cargo requirements need their own assessment. Explain the complete shipment scope so the agent can distinguish the supported inland work from any additional service requiring review.

Follow a move from pricing to a shipment update

The following example illustrates how to organize requests. It is not a live quote, booking, or guarantee of a particular price or transit time.

Request a quote and understand its scope

Start with the APM Elizabeth-to-Linden request above. The agent needs to identify the intended terminal and facility location, preserve the supplied equipment and cargo facts, and resolve any other information required by the available operation.

If your shipment has a special operating requirement, include it before pricing. For example, a planned overnight container drop is relevant to the work being quoted. Adding it only after receiving a price can change the required operation and its cost.

Once the quote returns, ask:

Explain which movements and services this quote covers, which quantities or conditions were assumed, and what could create an additional charge.

Check the following points before moving on:

• The quote describes the correct import or export movement.
• The terminal and facility are the intended locations.
• The equipment, net cargo weight, and declared services match your request.
• The total's currency and pricing unit are clear.
• Empty-container and chassis movements are understood.
• The quote's terms, validity, and unresolved conditions are visible.

Keep the quote reference. If you need the quote again, ask the agent to retrieve that reference. If shipment facts change, identify the changes and request the appropriate revised pricing rather than assuming the earlier amount still applies.

Review the booking information and authorize the move

You can ask the agent to prepare the next step without creating a booking:

Tell me what information is still needed to book this quote. Do not create the booking yet.

Required booking information includes the selected quote number, facility name, street address, city, state, ZIP code, facility contact email, and container number. Supply the actual details for the move and check them before submission.

When you are ready, make the instruction explicit:

Book the selected quote using the facility and container details I provided.

The agent should report the returned booking outcome and shipment reference. Read the result at the level it establishes. A created shipment record, carrier assignment, a terminal appointment, and an email delivery are different events. Only treat each as confirmed when the relevant result or shipment record confirms it.

If the response is interrupted, establish whether the booking was created before asking for another submission. Repeating a consequential request without checking its outcome can create confusion even when recovery controls exist.

Track progress and prepare a useful update

Later, ask the agent to retrieve the shipment using its returned reference:

Check this shipment again. What is the latest completed milestone, what happens next, and what ETA is available for the next stop? Include the relevant observation times.

An operational answer should let you separate:

• Recorded events: What the shipment record says has happened.
• Planned events: Appointments or other intended steps.
• Estimates: An available arrival prediction and the stop it concerns.
• Missing facts: Information that has not been recorded or cannot currently be established.

For example, a delivery appointment with no actual arrival timestamp establishes the appointment. It does not establish that the truck arrived. Similarly, a recently written response can describe an older vehicle position. Ask when the observation was recorded if freshness matters.

After reviewing the facts, you can ask:

Draft a short customer update from this tracking result. Include the next step and any uncertainty. Leave it here for review.

Drafting the message uses the agent's writing ability. Sending it through email or another system requires that system's own available tools and authorization. If the question concerns recorded charges, the accessorial charge guide explains how to inspect amounts, reasons, and supporting evidence.

How MCP fits alongside Portmate chat and APIs

Different interfaces suit different working patterns. MCP adds a standard way for agents to use exposed capabilities, while the business services underneath continue to perform their work.

An agent can use an API through a custom integration, and an MCP server can call APIs or domain services behind its tools. The distinction is how the integration presents capabilities to the agent. MCP standardizes tool discovery and invocation; it does not remove the need for schemas, permissions, business rules, or software maintenance.

Microsoft's integration guidance discusses when MCP is useful compared with connectors or direct API calls. A shared tool interface can help when several agents need the same capabilities. A direct API may suit an existing application with a well-defined integration path.

For a logistics team, the choice is practical. Use the interface that fits the job, the systems you already operate, and the controls your organization requires. You can continue using a product interface for detailed inspection even when an agent handles the initial request or explanation.

Bring your own agent and request preview access

Bringing your own agent means using an agent application you choose to interact with Portmate's exposed tools. It does not require moving the model itself into Portmate or giving the agent unrestricted access to your other systems.

Check the application hosting your agent

MCP support belongs to a specific application and configuration. A model name or familiar assistant logo alone does not tell you which connections that application can use.

Official documentation describes MCP connection capabilities in several environments:

• ChatGPT: Custom MCP servers can expose read and write tools through a plugin, subject to workspace permissions and security restrictions.
• Codex: External MCP connections provide tools and information from systems outside the repository.
• Claude: Remote MCP connectors connect the application to internet-hosted tools and information.
• Gemini CLI: Configured MCP servers let the CLI discover and call external tools.

These are examples of applications with documented MCP capabilities. For Portmate preview access, confirm the supported setup for the particular application, transport, and authentication method you intend to use. An organization's administrator may also control which connections and actions are permitted.

A company can build its own agent application around an MCP client as well. The protocol's official SDKs provide implementation resources for development teams.

Start with one defined workflow

Use the preview to establish a useful, repeatable task before combining several systems or automating consequential actions.

1. Request access. Contact support@portmate.ai or use Portmate's contact options. Include your company, the agent application you want to use, and whether you want to begin with quoting, booking, tracking, or another specific workflow. Ask for the current preview connection instructions and requirements.

2. Follow the supplied connection instructions. Use the server details and authentication flow provided for your access. Configuration differs between agent applications, so use the instructions for your actual environment rather than adapting an unrelated server's example.

3. Review the available tools. Establish what the connection exposes and what information its operations require. Start with a lookup for a shipment you are authorized to access, or review the tool catalog before attempting an operation that creates a record.

4. Try one task and inspect the outcome. For tracking, compare the returned reference, milestone, and timestamps with the shipment record. For quoting, review the scope and assumptions. For booking, verify the selected quote and required details before authorizing submission.

If your account needs attention, the guides to signing in and creating an account explain the ordinary Portmate account workflows. Use the preview instructions for the external connection's own authentication requirements.

Keep permissions and results clear

An agent connection becomes useful when the team can understand both what it is allowed to do and what actually happened. Establish those boundaries around the operation you are performing.

Distinguish reading from creating or sending. Looking up tracking information is different from creating a quote, creating a booking, or sending a copy. A quote request can create a business record even if you are still evaluating the move. State whether you want an explanation, a lookup, a new quote, or a booking.

Make consequential instructions specific. Reviewing a quote does not itself request a booking. Knowing the facility address does not establish permission to create a shipment. If several quotes or containers are in the conversation, identify the one you want the agent to act on.

Use the connected service's results as evidence. An agent should explain the amounts, references, assumptions, and status returned by Portmate. It should not fill gaps with a plausible price, unrecorded milestone, or invented confirmation. Treat the result of a partial or failed request according to what is known about the actual operation.

Keep shipment facts aligned across steps. A natural conversation may spread the facts across several messages. Check that the agent preserves the active shipment's terminal, equipment, weight, and requested services when it constructs a later request. A correction for one container should not silently alter a different move.

Treat retrieved material as information. Documents, articles, and external text can help explain a shipment or supply facts. Instructions embedded in them do not authorize unrelated bookings, account actions, or disclosure of business information. State your task directly and review actions that affect records or communicate externally.

Use the provided authentication flow. Grant only the access appropriate to the workflow and follow your organization's connection policies. Authentication establishes access; it does not mean every possible action is desired. Your agent provider's data handling and retention settings also matter when shipment information enters that environment.

The MCP tools specification requires server-side validation and access controls and recommends user visibility and control over sensitive calls. The authorization specification describes the protocol's authorization framework. The controls present in your actual connection and agent application determine how those principles appear in use.

Resolve common workflow problems

A connection issue, missing input, and an unsupported service request need different remedies. Begin with the returned result and the task you were attempting.

For an unresolved problem, contact Portmate with the agent application, the operation attempted, the relevant shipment or quote reference, the error message, and the time of the attempt. Keep credentials out of the support message. A precise account of what happened helps the team distinguish setup, access, input, and execution problems.

Where agent-connected logistics could go next

MCP makes a broader working pattern possible: customers can bring an agent to a service instead of beginning every task inside that service's own interface. The service remains responsible for its capabilities and business rules; the agent can help the customer combine those capabilities into a useful sequence.

Consider a forwarder working from an incoming delivery request. With the appropriate connections and permissions, an agent could extract the supplied lane details, obtain drayage pricing, prepare a comparison, and draft an internal recommendation. After approval, it could continue with booking and later retrieve the shipment's recorded progress. Each connected system would still need to expose the necessary operation and authorize its use.

This direction already has examples in freight. Flexport's MCP offering describes customer-agent workflows for rates, shipment information, and bookings. That is evidence of the pattern appearing in logistics, rather than a promise that every provider, shipment type, or workflow is interoperable today.

Some future workflows may also involve separate agents working across company boundaries. MCP connects an agent to exposed tools and information. Agent-to-agent protocols address collaboration between agents. The official A2A comparison explains how those approaches can complement each other. An MCP server may expose sophisticated behavior, but using MCP alone does not establish that independent agents are negotiating a shipment.

The practical opportunity is to make a customer-approved task easier to carry across systems. That requires accurate shipment identities, useful tool descriptions, explicit permissions, clear results, and a way to resolve exceptions. Scheduling repeated checks, running in the background, or applying spending limits also requires an agent application or automation system that supports those controls. The connection alone does not supply them.

Questions to settle before your first integration

These decisions help turn an interesting demonstration into a workflow your team can use.

Do we need to write code? An agent application with a supported connection flow may let an end user configure access without building an integration. A custom company agent still needs an MCP client and application logic. Confirm the setup available for your Portmate preview access before estimating the work.

Can we keep using Portmate directly? MCP provides an additional interface. Use Portmate directly when you want its own chat, tracking workspace, or other product views. An agent answer can also be the starting point for inspecting the underlying shipment record.

Will any MCP-capable agent behave the same way? A common protocol helps applications connect, but models and hosts differ in tool selection, clarification, approvals, context handling, and presentation. Evaluate the actual agent application against the tasks your team needs.

What will the connection cost? Ask Portmate about the current preview terms and check your agent provider's subscription or usage charges separately. MCP is a protocol, so its name alone does not establish the commercial terms of either service.

Can we start with tracking and add booking later? Begin with the tools and permissions available to your connection. A narrow initial task is easier to inspect and repeat. Before expanding into booking or cross-system actions, decide who can authorize them and how the team will verify their outcomes.

For preview access, send the Portmate team a concrete starting request: the agent you want to connect, the workflow you want it to perform, and the people who will review its actions. That gives the first connection a clear purpose and makes its results easier to evaluate.