MCP API

Overview

The TeamCentral MCP (Model Context Protocol) API enables AI assistants — such as Claude, ChatGPT, and Microsoft Copilot — to interact directly with your TeamCentral data through a standardized protocol. Once connected, an AI assistant can query your data models, look up entities, execute saved queries, extract file content, and submit data entry forms on behalf of authenticated users.

The MCP API acts as a bridge between AI platforms and your existing TeamCentral Platform API, translating natural language interactions into structured API calls while enforcing the same authentication and authorization policies that govern your platform.

What is MCP?

The Model Context Protocol is an open standard that allows AI models to discover and use external tools and data sources. TeamCentral implements MCP so that AI assistants can securely access your integration data without custom per-platform development.

Connecting to TeamCentral

Claude

To connect Claude to your TeamCentral instance:

1. Open Claude.ai, Claude Desktop, or Claude Code
2. Navigate to the MCP server settings and add a new remote server
3. Enter your TeamCentral MCP API URL (provided by your administrator)
4. Complete the OAuth sign-in flow — you will be redirected to Microsoft Entra to authenticate with your organizational credentials
5. Grant consent for the MCP server to access TeamCentral on your behalf

Once connected, Claude can use all available TeamCentral tools in your conversations.

ChatGPT

To connect ChatGPT to your TeamCentral instance:

1. Open ChatGPT and navigate to settings
2. Add a new MCP connection and enter your TeamCentral MCP API URL
3. Complete the OAuth sign-in flow through Microsoft Entra
4. Grant consent for the MCP server to access TeamCentral on your behalf

Microsoft Copilot

For Microsoft Copilot integrations (Copilot Studio, M365 Copilot), contact your TeamCentral administrator. These platforms use a pre-configured connector and may require additional setup in your Entra tenant.

Authentication

The MCP API uses OAuth 2.0 with Microsoft Entra ID for authentication. When you connect an AI assistant to TeamCentral, the following occurs:

1. Authorization — You sign in with your Microsoft Entra credentials
2. Consent — You approve the MCP server to access TeamCentral on your behalf. A single consent prompt provisions both the MCP API and Platform API in your tenant
3. Token exchange — The MCP API exchanges your token for a Platform API token using the On-Behalf-Of (OBO) flow
4. Scoped access — All data access is limited to what your user account has permission to see in TeamCentral

Important

The MCP API enforces the same permissions as the TeamCentral platform. Users can only access data models, entities, and saved queries that their account has been granted access to.

Available Tools

The following tools are available to AI assistants through the MCP API. Each tool summary is shown below — for complete input schemas, parameter details, and usage notes, see the MCP Tool Reference.

Data Models

Tool	Description	Type
get_data_models	Returns all Data Models configured in your environment	Read-only
get_data_model_properties	Returns the properties (fields) of a specific Data Model	Read-only

These tools allow an AI assistant to understand the structure of your data — what models exist and what fields they contain — before querying or submitting data.

Entity Lookup & Query

Tool	Description	Type
lookup_by_keyword	Searches for entities by keyword across related entity types	Read-only
query_entities	Queries entities with filters, sorting, pagination, and optional property selection	Read-only
get_entities_by_ids	Retrieves one or more entities by their unique IDs	Read-only

Query Operators

When using query_entities, the following filter operators are available:

Operator	Description
EqualTo	Exact match
NotEqualTo	Not equal
GreaterThan	Greater than
GreaterThanEqualTo	Greater than or equal
LessThan	Less than
LessThanEqualTo	Less than or equal
Contains	Contains substring
StartsWith	Starts with
NotStartsWith	Does not start with
EndsWith	Ends with
NotEndsWith	Does not end with

Saved Queries

Tool	Description	Type
get_saved_queries	Lists all Saved Queries (Live Data Providers) accessible to the user	Read-only
get_saved_query_details	Returns metadata and mapped fields for a Saved Query	Read-only
execute_saved_query	Executes a Saved Query with optional input parameters	Read-only

Saved Queries allow administrators to define pre-built queries with parameters that AI assistants can execute. This is useful for common lookups like "show me all open opportunities" or "find employees in a specific department."

Data Entry

Tool	Description	Type
get_data_entry_form	Retrieves a data entry form for a Data Model, optionally pre-populated	Read-only
submit_data_entry_form	Submits a completed data entry form to create or update a record	Destructive

TIP

The get_data_entry_form tool can return forms in JSON or Adaptive Card format. JSON format provides field metadata for conversational data entry, while Adaptive Cards render rich forms in platforms that support them (e.g. Microsoft Teams).

File Extraction

Tool	Description	Type
get_file_text	Extracts readable text from document URLs (PDF, DOCX, PPTX, HTML, and more)	Read-only

When query results include File entities with an externalFileLocation property, use get_file_text to extract the document's text content. The tool returns text in chunks — call repeatedly using the nextOffset value until hasMore is false.

Example Interactions

Here are examples of how an AI assistant might use these tools in a conversation:

Exploring available data:

"What data models are available in TeamCentral?"

The assistant calls get_data_models and presents a list of your configured models.

Looking up a record:

"Find me the customer record for Contoso"

The assistant calls lookup_by_keyword with "Contoso" to find matching entities.

Running a report:

"Show me all open purchase orders over $10,000"

The assistant calls get_saved_queries to find a relevant query, then execute_saved_query with the appropriate parameters.

Submitting data:

"Create a new support ticket for Contoso about a billing issue"

The assistant calls get_data_entry_form to get the form structure, fills in the values from your request, and calls submit_data_entry_form to create the record.

Reading a document:

"What does the attached contract say about payment terms?"

The assistant retrieves the file entity, then calls get_file_text with the document URL to extract and analyze the text content.

Administrator Setup

For administrators deploying the MCP API for their organization, the following setup is required:

• Entra app registration and configuration
• Environment variables and deployment
• Connector Security tab setup

Note

The MCP API requires three Entra app registrations (Connector, MCP API, and Central Platform API) and an Azure Storage account for OAuth state management. Once deployed, users can connect their AI assistant without additional admin setup — OAuth client registration is handled automatically. Contact your TeamCentral representative for detailed setup instructions.

Troubleshooting

Common Issues

"Unauthorized" or sign-in fails

• Verify your Entra credentials are correct and your account has access to TeamCentral
• Check that the MCP API's Entra app registration is properly configured
• Ensure you have completed the consent prompt when first connecting

No data models or entities returned

• Confirm your user account has the appropriate roles and permissions in TeamCentral
• Verify that Data Models and Connectors are configured and active in your environment

Form submission fails

• Check that all required fields are populated
• Review the error message — the MCP API returns field-level validation errors when available

Connection timeout or slow responses

• Large query results are automatically truncated to stay within platform limits
• Use filters and the properties parameter to narrow results when querying large datasets
• For paginated queries, use smaller pageSize values to reduce individual response size