Port AI

Port AI is the foundational AI system that enables intelligent interaction with your Port data through natural language. As an MCP (Model Context Protocol) client, Port AI accepts prompts and tools, runs autonomous processes to query your software catalog, and returns comprehensive responses.

What is Port AI?

Port AI serves as the base interface that:

• Accepts prompts: Receives natural language queries and requests.
• Uses developer MCP tools: Uses developer MCP tools from the Port MCP server to access your catalog data. Port AI currently supports developer tools (for querying and running actions) but not administrative tools (such as creating blueprints or managing scorecards).
• Runs autonomous processes: Intelligently determines which tools to use and how to combine them.
• Returns responses: Provides comprehensive answers and can execute actions based on your requirements.

Port AI acts as an MCP client that connects to Port's remote MCP server, giving it access to your entire software catalog, blueprints, entities, and configured actions.

How Port AI works

When you interact with Port AI:

1. You provide a prompt - Ask questions or request actions in natural language.
2. Port AI analyzes your request - Determines what information or actions are needed.
3. Tools are executed - Port AI uses MCP server tools to query your catalog, analyze data, or prepare actions.
4. You receive a response - Get comprehensive answers, insights, or action execution results.

Where to use Port AI

Interface	Best for	Documentation
Port AI Assistant	Quick questions in Port with no setup	Port AI Assistant
AI chat widget	Embedded chat on dashboards	AI chat widget
Slack app	Team chat and notifications	Slack app
Port MCP server	IDE and external agent access	Port MCP server
AI workflows	AI steps inside Port workflows	AI action node
Port n8n node	n8n automations against the Context Lake	Port n8n node
MCP connectors	Governed access to external MCP servers	MCP connectors

Port AI is grounded in the Context Lake. The more your catalog is modeled, the more accurate AI responses become.

Supported capabilities

Tools

Port AI leverages developer tools from the Port MCP server to query your catalog, analyze scorecards, and execute self-service actions.

Examples:

• "Show me all microservices owned by the Backend team" → Uses data query tools to search entities.
• "Create a new incident report for the payment service" → Uses action execution tools to run an incident creation action.
• "What's the security score for our production services?" → Uses data query tools to access scorecard information.

For tool execution modes, approval settings, and API configuration, see API interaction.

Skills

Skills are domain-specific guidance packages that help Port AI handle specialized tasks more effectively. Port includes built-in skills for common workflows, and you can create custom skills tailored to your organization.

Learn more about skills →

Memory

Memory allows Port AI to learn and remember your preferences across conversations - output formats, coding styles, and common workflows. Memory is enabled by default and can be managed via API.

Learn more about memory →

Common use cases

Port AI excels at helping platform engineers and developers with:

Information discovery:

• "Who is the owner of service X?"
• "How many services do we have in production?"
• "Show me all microservices owned by the Backend team".
• "What are the dependencies of the OrderProcessing service?"

Quality and compliance analysis:

• "Which services are failing our security requirements scorecard?"
• "What's preventing the InventoryService from reaching Gold level?"
• "Show me the bug count vs. test coverage for all Java microservices".

Running actions:

• "Can you help me deploy service X to production?"
• "Create a new incident report for the payment service outage".
• "Set up a new microservice using our standard template".
• "Notify the reviewers of pull request #1234".

LLM models and providers

Port AI uses state-of-the-art Large Language Models to power all AI interactions:

• Port's managed AI infrastructure (default) - enterprise-grade security with automatic updates.
• Bring your own LLM - configure your own providers for data privacy, cost optimization, and compliance.

For comprehensive information about LLM provider management, see LLM provider management.

Limits and usage

Port AI operates with per-minute rate limits and a monthly invocation quota at the organization level. When using Port's managed AI infrastructure, the default quota is 500 AI invocations per month. Bring your own LLM provider to use your provider's limits instead.

For rate limits, quota monitoring, error handling, and requesting increases, see Rate limits and quotas.

Security and governance

Port AI respects your organization's RBAC, data access policies, and audit requirements. All interactions are logged, and Port AI can only access data you have permission to view.

For access controls, data retention, and admin policies, see AI security and data controls.

Getting started

• Quick answers in Port - start with the Port AI Assistant.
• Programmatic integration - use API interaction.
• IDE and external agents - set up the Port MCP server.
• Custom automation - build Port custom agents.

Frequently asked questions

What are the main use cases Port AI supports? (Click to expand)

Port AI supports two primary interaction types:

1. Ask me anything (information queries)

   • Natural language queries about your development ecosystem.
   • Examples: "Who owns service X?", "What's the deployment frequency of team Y?"
   • Focused on surfacing information from connected data sources.

2. Run an action (task assistance)

   • Assist with running or pre-filling self-service actions.
   • Examples: "Create a bug report", "Set up a new service".
   • You can decide whether Port AI runs the action automatically or requires approval.

How can I interact with Port AI? (Click to expand)

See Where to use Port AI for the full list of interfaces, including the Port AI Assistant, AI chat widget, Slack app, API, workflows, and MCP server.

What happens if Port AI can't answer my question? (Click to expand)

If Port AI doesn't have access to the data needed to answer your question, you'll receive a response indicating that it can't assist with your query. This typically happens when:

• The requested data isn't available in your Port catalog.
• You don't have permissions to access the relevant data.
• The question is outside the scope of Port AI's capabilities.

Consider rephrasing your question or checking your data permissions.

Can I customize how Port AI responds? (Click to expand)

Yes, you can customize Port AI in several ways:

• Use the AI chat widget with custom prompts and conversation starters.
• Build AI agents for specialized use cases.

All customizations operate within Port's secure framework and governance controls.

Which LLM models does Port AI use? (Click to expand)

Port AI uses advanced language models depending on the interface and configuration. By default, Port's managed infrastructure uses models such as GPT-5 and Claude Sonnet, Haiku, and Opus. You can also bring your own LLM from providers including OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, and others.

Learn more about LLM provider management.

How is my data handled with Port AI? (Click to expand)

Port AI is designed with security and privacy as priorities. All data processing occurs within Port's secure cloud infrastructure, your data is not used for model training, and interaction data is stored for up to 30 days for operational purposes.

For comprehensive security and data governance information, see AI security and data controls.

What if Port AI gives incorrect answers? (Click to expand)

Port AI can make mistakes. If you're receiving incorrect answers:

1. Check what tools and data sources were used in the AI invocation record.
2. Verify that the relevant data is correctly configured in your Port catalog.
3. Try rephrasing your question or breaking it into smaller, more specific queries.
4. Contact support if problems persist.