> For the complete documentation index, see [llms.txt](https://docs.frends.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.frends.com/frends-development/agentic-ai/frends-mcp.md). # Frends MCP Let AI use your Frends Tenant. Frends MCP is a Model Context Protocol server integrated directly into Frends, providing AI clients a controlled way to interact with the platform. Instead of just invoking external integrations, AI clients can use this feature to understand Tenant resources, manage Process drafts, and inspect operational history. This enables a more collaborative development and maintenance flow between human users and AI assistants. The feature is designed for integration development assistance, architectural reviews, and operational investigations. It provides a structured interface that allows AI clients to interact with Frends concepts using the platform's native logic and security rules. While it does not replace the Management Portal, it offers a powerful way to augment existing workflows with AI-driven insights and automation. ## Interacting with Frends via AI Frends MCP enables AI clients to work with Frends as a native protocol participant. This connection allows the client to retrieve information from a Tenant, browse available building blocks, and perform specific actions based on the permissions assigned to its access token. This shifts the AI's role from a simple consumer of APIs to an assistant that understands the context of the Frends environment. Commonly, this is used for assisted development where an AI client inspects existing integrations to propose new Process drafts. By reading available Tasks and their requirements, the AI can help construct valid workflows that follow organizational standards. It can also assist during maintenance by reading Process Instance history to explain execution paths or identify why a specific run behaved in a certain way. ## Comparison with MCP Triggers Frends provides different ways to use the Model Context Protocol, and it is important to distinguish between Frends MCP server and the MCP Trigger. Frends MCP works at the platform level, allowing an AI to manage the Tenant, its Processes, and its deployment lifecycle. In contrast, an MCP Trigger is used within a specific Process to make that single integration callable as a tool. Choose Frends MCP when the goal is platform-level assistance, such as drafting Processes, exploring Task libraries, or performing forensic analysis on Instance history. The MCP Trigger is the correct choice when an AI simply needs to trigger a specific integration or fetch data through an existing Process. Combining both features allows organizations to benefit from AI both in how they build integrations and how they use them. ## Frends MCP capabilities Frends MCP provides a comprehensive set of tools that allow AI clients to participate in the full integration lifecycle. These capabilities range from the initial design of a Process to the investigation of runtime behavior in production environments. ### Process authoring and validation AI clients can assist in authoring by creating Process drafts and adding elements like Triggers, Tasks, and flow control logic. Authoring always occurs within a draft, which remains separate from active Process versions. This allows for iterative building and validation without affecting any live integrations or requiring immediate deployment. Once a draft is complete, it can be saved as a Process version and deployed to a development Agent Group. This flow ensures that AI-generated content follows a standard development path where a human developer can review the final result. Certain configurations, such as Process Variables and API Policies, still require manual setup in the Frends Management Portal to ensure full security and functional control. ### Tenant and metadata exploration To provide accurate assistance, Frends MCP allows AI clients to explore the contents of a Tenant. Clients can list Processes, inspect their internal structures, and browse the library of available Tasks along with their parameter schemas. Access to Environment Variables is also possible, providing the AI with the necessary context to understand how different systems are connected. Exploration can be targeted using filters such as name, Agent Group, or Trigger type. While the server handles primary filtering, additional metadata like tags are returned to allow the AI client to further refine its understanding. This transparency ensures the AI has a clear view of the existing integration landscape before proposing changes or explaining existing logic. ### Operational visibility and deployment Beyond Process design and development, Frends MCP offers tools for managing the Frends runtime environment. AI clients can be granted permission to deploy Processes to Agent Groups and start Process runs. This is often combined with visibility into Process Instance history, where the AI can retrieve details about past executions to help diagnose errors or explain complex data flows. Operational actions are strictly controlled through scopes, meaning visibility does not automatically imply the power to deploy or execute. This granular control allows organizations to use AI for troubleshooting and observation without necessarily granting it the power to modify the production state. ## AI guides and platform skills Frends MCP includes built-in guides that instruct AI clients on how to perform complex sequences of actions. These guides go beyond simple tool descriptions by explaining the correct order of operations for tasks like diagnosing a failure or authoring a new integration. Following these guides ensures that the AI's actions remain consistent with Frends' best practices. In addition to guides, Frends provides skills that teach AI clients how to use the available tools effectively. An AI client can discover these resources to improve its performance and reliability when interacting with the platform. This structured learning approach helps bridge the gap between a general-purpose AI and the specific requirements of the Frends integration engine. ## Security and access control Security for Frends MCP is managed through Platform Applications and tokens. Administrators or users can create applications within the Frends UI to issue the tokens required for AI clients to connect. Personal applications inherit the permissions of the creating user, while Tenant-wide applications provide a more centralized way to manage broader access. Every token is defined by its expiry and a set of scopes that strictly limit what the AI client can see and do. The Frends MCP server enforces these scopes at the protocol level, only presenting the tools that the token is authorized to use. This ensures that the AI client operates within a "least privilege" model, protecting sensitive data and critical integration infrastructure. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.frends.com/frends-development/agentic-ai/frends-mcp.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.