What is an MCP? Model Context Protocol is a standard and secure way for an AI system to access other software.
How does MCP control user access and keep data safe? To ensure your data is secure, when you connect your AI system to some software’s MCP, you log in with the same login credentials you normally would use to log into that software. MCP uses the OAuth 2 standard. As you log in as you when an AI is using an MCP connection to talk to software, the software will automatically limit the AI’s access to the normal permissions linked with that user.
Can I set up an MCP integration myself, or do I need an IT administrator? At most companies with enterprise AI systems shared across the company, only IT administrators will have access to enable MCP servers. For most AI systems, this is true whether you are trying to enable a marketplace integration or add a custom MCP connection.
Configuration and Access Control
What AI systems does the Lattice MCP Server support? You can create a custom connection with the Lattice MCP Server and any AI system that supports the MCP protocol.
What is a Redirect or Callback URL and where do I find it? When you configure a custom MCP Server in Lattice, you must provide a Redirect URL that is specific to the AI system you are connecting to. Your AI system's administrator will provide you with this URL. This URL is used for authentication: when you log into Lattice, Lattice uses this Redirect URL to securely send a temporary access code back to your AI system. This code tells your AI system that you have successfully logged in and granted permission to connect. Note: When you are first connecting your AI system to the Lattice MCP Server, as a common error is due to entering an incorrect Redirect URL in Lattice. If this occurs, the AI system administrator will be able to save the connection to the Lattice MCP Server, but you will receive an authentication error in your AI system when you attempt to log in.
Does Lattice MCP Server support dynamic client registration (DCR)? Lattice MCP Server does not support DCR. Dynamic client registration lets an AI system automatically register itself as an OAuth client with the product, without an administrator manually creating a configuration first. In Lattice, you must explicitly create an MCP Server to connect to a new system.
How do I know if my AI system needs a client secret? Your AI system administrator will tell you if they need a client secret to connect to the Lattice MCP Server. In most cases where your AI system is operating on an end user's laptop or on their web browser, it will not need a client secret, only the Lattice MCP Server URL and Client ID. This type of connection is called Proof key for code exchange (PKCE). A client secret is more likely to be needed if the AI system you are connecting to Lattice is used for centralized automation, communicates with Lattice via an AI gateway or MCP proxy, or runs in a secure cloud execution environment.
Why can I see my client secret only once? Client secrets must be kept confidential for security reasons, and a common software security practice is to make a key visible only at the time of creation. Lattice allows you to generate a new client secret for an MCP Server, which will sever the connection to any AI system using the prior client secret. This is typically done to intentionally sever existing connections if a system has been compromised, but may also be necessary if the AI system you are connecting to is making certain changes, upgrades, or migrations.
How do I share my client secret? It is vital to your system security to treat a client secret carefully. Do not paste your client secret into a document or directly into an email or chat. Instead, store your client secret in an encrypted password vault or one-time link service. Your IT and Security team will be able to advise you on your company's recommended tools for sharing client secrets with your AI system administrator or integration partner.
How do I know if my Lattice MCP Server is successfully connected to my AI system? First, you will save your Lattice custom MCP Server in the Lattice Admin platorm, and your AI system administrator saves the Lattice MCP Server in the AI system and grants you access. When you log into your AI system, you will need to take additional steps to initiate the connection, and authenticate. If this is successful, try prompting with 'Who am I in Lattice?' in your AI system, and if properly connected the AI system will call the Lattice MCP and respond with your name.
How do I troubleshoot an authentication error? If you attempt to authenticate into a new connection to the Lattice MCP Server and cannot successfully authenticate, you must check that all of the correct information is entered in the Lattice MCP Server configuration in the Lattice admin platform, as well as the MCP configuration in the AI system. The most common issues are an incorrect redirect URL in the Lattice MCP Server configuration, or an incorrect MCP Server URL or Client ID in the AI system MCP configuration.
Using the Lattice MCP
What can I do with the Lattice MCP? Lattice MCP supportsthese tools and prompts. Currently, the Lattice MCP is designed to support manager and employee tasks.The MCP does not provide access to data or capabilities that you view in the Admin section of the Lattice platform.
How do I use the Lattice MCP in my AI assistant? In your prompts with the AI assistant, describe the task that you want the AI assistant to accomplish in natural language, and the AI assistant will call the appropriate Lattice MCP tools to complete the task. If you have multiple MCPs connected to your AI assistant within the same domain, it helps to specify explicitly if you want the AI assistant to collect data from or perform a task in Lattice to prevent confusion. For instance, if you ask your AI assistant to list your direct reports and you are connected to both the Lattice MCP and the MCP of your HRIS, your AI assistant may not know which to call, so you will get a more predictable response if you are explicit.
Does using the Lattice MCP consume tokens in my AI system? Yes, using any MCP system in your AI system will consume tokens, though the amount and any billing implications will depend on your AI system.
How do I see Lattice MCP’s full response? What you see in your chat window is not the Lattice MCP’s direct response to your AI system, but rather your AI system’s interpretation of that data converted to a user friendly format, such as human readable prose or a table. Based on your prompt, the AI system may show only a subset of the returned data, or possibly misunderstand the returned data. You can use prompts such as “show me the raw data returned by the MCP in the last prompt” to see the actual data returned in JSON format
How do I automate a workflow in Lattice using the MCP? This depends on the specific capabilities of your AI assistants. In general, there are two important concepts to understand: pre-defined workflows, and agentic capabilities. AI assistants have their own systems and structures for pre-defined workflows, such as Claude Skills or ChatGPT Custom GPTs. These allow you to take a workflow that you have just completed in your AI assistant and save that workflow so it can be run repeatedly or shared with others. Agentic capabilities allow an AI assistant to initiate a workflow without you directly prompting it each time. This can include scheduled workflows, or an AI workflow triggered by a change in another system. Learn more in Lattice MCP Suggested Prompts.
