TrackonAI MCP Server Documentation

TrackonAI MCP Server

The TrackonAI MCP Server connects AI assistants to your live commodity trading data using the Model Context Protocol. Instead of switching between screens and running manual reports, traders can ask questions in plain language and get answers from real data.

The server gives read access to your TrackonAI platform, covering contracts, shipments, open positions, invoices, counterparties, warehouse inventory, shipment documents, and analytics reports. Users with write access can also create and update records, such as new contracts, and generate trade documents. Every change goes through the same validation and approval workflows as the TrackonAI web application.

Getting Started

Claude.ai

Go to Settings > Integrations > Add MCP Server
Enter the server URL: https://mcp.trackon.ai/mcp
Sign in with your TrackonAI credentials when prompted
Start asking questions about your trading data

Claude Code

claude mcp add --transport http trackonai-mcp https://mcp.trackon.ai/mcp

Other MCP Clients

Any MCP-compatible client can connect using the server URL above. OAuth 2.1 authentication is handled automatically through the standard MCP auth flow.

Prerequisites

An active TrackonAI account
Your tenant must have MCP access enabled by an administrator
Your user must be added to the MCP access group
To create or change records, your user must be added to the MCP write access group

Contact your account administrator or support@trackonsoftware.com to enable MCP access for your tenant.

Available Tools

Read tools are available to every MCP user. Write tools require MCP write access.

Tool	Access	Description
`search`	Read	Search across all entity types by describing what you are looking for
`get`	Read	Retrieve full details of any entity by its ID
`list_entities`	Read	Query entities with structured filters and sorting
`get_entity_schema`	Read	Inspect field definitions and data types for any entity
`get_document_url`	Read	Generate a download link for shipment documents (B/L, COA, packing lists)
`list_analytics_queries`	Read	Discover available analytics reports and their parameters
`run_analytics_query`	Read	Execute reports (P&L, AR/AP aging, inventory, sales/purchase summaries)
`get_query_results`	Read	Poll report status and get CSV download link when complete
`get_parameter_values`	Read	Get valid dropdown values for report parameters
`validate_contract_for_approval`	Read	Check whether a contract can be sent for approval and list every problem to fix
`create_entity`	Write	Create a record, such as a contract, its positions and ports
`update_entity`	Write	Change fields on an existing record, or send a contract for approval
`delete_entity`	Write	Delete a record
`generate_document`	Write	Generate a trade document (Shipping Instructions, Proforma Invoice, Contract) from a template
`save_direct_posting`	Write	Prepare a manual ledger entry: debit or credit note, payment, receipt, or a transfer between accounts
`activate_direct_posting`	Write	Post a prepared ledger entry, or send it for approval where your company requires one

What You Can Ask

Here are examples of questions you can ask through any connected AI assistant:

Positions & Inventory

"How much Gedaref sesame seeds do I have left?"
"Show me all open purchase positions for sunflower oil"
"What is the total value of inventory in the Dubai warehouse?"

Shipments

"What shipments are arriving this week?"
"Show me the ETA for all vessels heading to Rotterdam"
"Get me the Bill of Lading for the shipment on MV Ocean Star"

Contracts

"List all open purchase contracts for butter"
"Show me contracts expiring this month"
"What is the total contracted quantity with Vendor A?"

Creating Contracts (write access)

"Create a purchase contract for 500 MT of sesame from Vendor A, shipment in May"
"Can this contract be sent for approval? What is missing?"
"Send contract 1042 for approval"

A new contract is always created as a draft. Before sending it for approval, the AI assistant runs the same checks as the web application and lists anything that needs fixing. Approval itself stays with your approvers: the assistant can send a contract for approval, but it cannot approve it.

Accounting Entries (write access)

"Record a debit note of 2,500 USD to Vendor A for the short delivery"
"Move the 10,000 USD Customer A owes us to Customer B"

The AI assistant shows you the entry before posting it. Debits always equal credits, closed periods stay closed, and where your company uses approvals, the entry waits for an approver.

Financial

"Who owes us money and how many days overdue?"
"Run the AR aging report for this quarter"
"Show me the P&L report by trader for March"

Documents

"Find the Certificate of Analysis for shipment JRN-2026-0142"
"Download the packing list for the latest butter shipment"

Entity Types

The server provides access to 60+ entity types. The most commonly queried are:

Entity	What It Represents
`JOURNEY`	Shipments, voyages, vessel movements, ETAs
`POSITION`	Open buy/sell positions, remaining quantities, exposure
`CONTRACT`	Purchase and sale agreements
`INVOICE`	Financial documents, payments, overdue amounts
`SUB_ACCOUNT`	Counterparties, trading partners (buyers, sellers, banks)
`PRODUCT`	Commodity definitions (butter, sesame, sunflower oil)
`SUB_DEPARTMENT`	Warehouses and stock locations (including In Transit)
`JOURNEY_DOCUMENT`	Uploaded shipment documents (B/L, COA, packing lists, weight certificates)

Use the get_entity_schema tool to discover all available fields, data types, and filter options for any entity type.

Filtering & Sorting

The list_entities tool supports structured filters using these operators:

EQ: equals
NE: not equals
GT / GTE: greater than / greater than or equal
LT / LTE: less than / less than or equal
CONTAINS: partial text match
BEGINS_WITH: prefix match

You do not need to construct filters manually. Simply describe what you are looking for and the AI assistant will build the appropriate query.

Analytics Reports

Pre-built analytics reports are available for common trading analyses:

Shipment P&L
Accounts Receivable / Payable Aging
Inventory Aging
Sales by Product
Purchase by Product
Open Contract Details
Journey Report

Reports run asynchronously. The AI assistant will automatically poll for results and provide a CSV download link when the report is complete.

Security

Authentication: OAuth 2.1 with PKCE via Amazon Cognito. The same credentials you use for the TrackonAI web application.
Tenant isolation: Your data is completely isolated. The server enforces the same multi-tenant security as the TrackonAI web application.
Opt-in access: MCP must be explicitly enabled per tenant and per user by an administrator. Tenants that have not opted in are invisible to MCP.
Read and write access: Every MCP user can query data. Creating, changing, or deleting records requires a separate write access group, granted per user by an administrator.
Same rules as the web application: Every change passes the same validation and approval workflows. Approvals, accounting postings, and other workflow steps cannot be performed directly through MCP.
Encryption: All communication uses HTTPS/TLS.

Data Privacy

TrackonAI itself is fully secure. Your data is isolated, encrypted, and inaccessible to our team. That does not change when you use AI assistants.

The AI assistant feature is optional. When enabled, you choose where the AI processes your data: on the AI provider's computers (simplest), in your own cloud account (private), or on your own hardware (maximum privacy).

See AI Assistant Privacy Options for a full comparison of each option and what it means for your data.

Troubleshooting

"No MCP-enabled tenants found"

Your tenant administrator needs to create an MCP access group and add your user to it. Contact your administrator or support@trackonsoftware.com.

"You have MCP access to multiple tenants"

If you have access to more than one tenant, tell the AI assistant which tenant you want to query (e.g., "Query the HFD tenant" or "Switch to the Mana account").

"You have read-only MCP access"

Creating or changing records requires MCP write access. Ask your administrator to add your user to the MCP write access group.

Authentication errors

Ensure you are signing in with the same credentials you use for the TrackonAI web application. If the issue persists, try disconnecting and reconnecting the MCP server.

Support

For questions, issues, or to request MCP access for your tenant:

Email: support@trackonsoftware.com
Website: trackon.com

Need a custom setup, or just have more questions about TrackonAI? Contact us for purchase, installation, or additional services.