Skip to main content

Harness AI Support Agent

Last updated on

The Harness AI Support Agent is your first line of support inside the Harness platform. It draws on Harness documentation and knowledge base content to answer product questions, offer troubleshooting guidance, and pull contextual examples straight from the official docs, so you get an answer in seconds instead of filing a ticket and waiting.

AI model

The Support Agent uses Claude Opus 4.6 to process your questions and retrieve relevant documentation.

What will you learn in this topic?​

By the end of this topic, you will be able to:

Before you begin​

Before you use the Support Agent, ensure you have the following:

  • Harness AI enabled: Harness AI must be active for your account. Go to Harness AI overview for more information on activating Harness AI.
  • Account settings access: Permission to view and edit default settings if you enable Harness AI yourself. Go to RBAC in Harness for more information on permissions.
  • License type: No specific license tier. The Support Agent is available to all Harness users with Harness AI enabled, regardless of license type.

Installation and setup​

The Support Agent ships as part of the Harness AI feature set, so there is nothing separate to install. You only need to turn Harness AI on.

  1. Navigate to Account Settings in the left navigation.
  2. Under General, select Default Settings.
  3. Find Harness AI and enable the Harness AI setting.
  4. Optional: Enable Allow Overrides to let individual Organizations and Projects override this setting.

Once enabled, the Support Agent is accessible through the Harness AI assistant interface.

Use the Support Agent​

Access the Support Agent through the Harness AI chat interface. The agent recognizes documentation-related questions on its own and retrieves the relevant content, so you do not need to route your question anywhere special.

Ask questions​

Ask the Support Agent anything about Harness products, features, or configurations. The agent searches the documentation knowledge base and returns the relevant information.

Example questions:

  • "How do I create a Kubernetes connector?"
  • "What are the required permissions for a GitHub connector?"
  • "Explain the difference between rolling and canary deployments"
  • "How do I configure failure strategies in a pipeline?"
  • "What environment variables are available in CI build steps?"

Get documentation examples​

The Support Agent retrieves specific examples from the documentation, including YAML configurations, API examples, and step-by-step guides.

Example prompts:

  • "Show me an example of a Kubernetes deployment manifest"
  • "Give me a sample pipeline YAML for a CI/CD workflow"
  • "What does a Terraform connector configuration look like?"
  • "Show me how to configure an approval stage"

Ask contextual follow-up questions​

The Support Agent keeps track of your conversation within a chat session, so you can ask follow-up questions without restating what you already told it.

Example conversation:

  1. You: "How do I set up a Docker connector?"
  2. Support Agent: Provides documentation on Docker connector setup
  3. You: "What permissions does it need?"
  4. Support Agent: Provides permissions information specific to Docker connectors, understanding the context from the previous question

Get troubleshooting guidance​

When you hit an error, the Support Agent helps you find the troubleshooting documentation that covers it.

Example prompts:

  • "I'm getting a 'delegate not found' error. How do I fix this?"
  • "My pipeline is failing with a connection timeout. What should I check?"
  • "How do I debug a failed Kubernetes deployment?"
  • "What are common causes for artifact download failures?"

Features​

The Support Agent combines natural language understanding with full documentation coverage and session memory. The following table summarizes what each capability gives you.

FeatureDescription
Natural language queriesAsk questions in plain English and receive clear, concise answers.
Documentation searchSearches across all Harness documentation to find relevant content.
Code examplesRetrieves YAML, JSON, and code examples directly from documentation.
Session contextMaintains conversation history for contextual follow-up questions.
Multi-module coverageAnswers questions across all Harness modules (CI, CD, CCM, STO, and others).

Natural language queries​

The Support Agent interprets your intent rather than matching keywords, so you do not need to know the exact terminology to find the right page.

How it works:

  1. You ask a question in natural language.
  2. The agent analyzes your question to understand the intent.
  3. The agent searches the documentation knowledge base.
  4. Relevant sections are retrieved and synthesized into a response.

The Support Agent supports the same languages as Harness AI, including English, Spanish, French, German, Hindi, Korean, Mandarin, Dutch and more. Go to Harness AI overview for more information on supported languages.

The Support Agent has access to the complete Harness documentation, including:

  • Platform documentation
  • Module-specific guides (CI, CD, CCM, STO, CE, FF, SRM, IDP, IACM)
  • API references
  • Tutorials and quickstarts
  • Troubleshooting guides
  • Best practices and reference architectures

The Support Agent pulls from the latest published Harness documentation. Documentation updates are reflected in Support Agent responses after they are published.

Code examples​

When documentation includes code samples, the Support Agent retrieves and displays them in your chat. This includes:

  • Pipeline YAML configurations
  • Connector definitions
  • Service and environment manifests
  • API request and response examples
  • Shell commands and scripts
  • Terraform and infrastructure code

Session context​

The Support Agent uses session memory to maintain context throughout your conversation. This enables:

  • Follow-up questions: Ask clarifying questions without restating context.
  • Progressive learning: Build understanding through a series of related questions.
  • Refinement: Ask the agent to elaborate or provide more specific examples.
  • Comparative questions: Ask how concepts relate to previously discussed topics.

Session context lasts for the duration of your chat session. Starting a new session clears the context, and each new session starts fresh without memory of previous conversations.

Best practices​

How you phrase a question has a direct effect on how useful the answer is. The following guidance helps you get precise responses instead of general overviews.

Write effective questions​

To get the most accurate responses from the Support Agent:

  • Be specific: Include relevant details like module names, resource types, or error messages.
  • Provide context: Mention what you are trying to accomplish.
  • Use follow-ups: Start broad and narrow down with follow-up questions.
  • Include error messages: When troubleshooting, include the exact error text.

Good example:

"How do I configure a GitHub connector with a personal access token for a private repository?"

Less effective example:

"How do connectors work?"

Use session context effectively​

Take advantage of session memory to have productive conversations:

  1. Start with the main topic: Ask your primary question first.
  2. Build incrementally: Ask follow-up questions that build on previous answers.
  3. Reference previous answers: Use phrases like "In that example..." or "For that connector...".
  4. Request elaboration: Ask "Can you explain more about..." when you need details.

Choose between the Support Agent and the DevOps Agent​

The two agents solve different problems. Use the Support Agent to learn and diagnose, and the DevOps Agent to build and change things.

Use Support AgentUse DevOps Agent
Learning about Harness featuresCreating pipelines, stages, or steps
Finding documentation and examplesModifying existing configurations
Troubleshooting guidanceGenerating services, environments, or connectors
Understanding best practicesAutomating DevOps tasks
Exploring configuration optionsBuilding and editing resources

Data privacy​

The Support Agent follows the same data privacy policies as all Harness AI features:

  • No training on your data: Your questions are not used to train AI models.
  • Ephemeral processing: Questions are processed in real time and not stored.
  • Documentation-only responses: Answers are sourced from public Harness documentation.
  • No account data access: The Support Agent cannot view your pipelines, configurations, secrets, or any account-specific data.

For more information, see Data storage and privacy policies.

Limitations​

The Support Agent answers questions based on Harness documentation. Knowing where that boundary sits tells you when to escalate instead of rephrasing.

  • Documentation scope: Can only answer questions covered in the official documentation.
  • No account access: Cannot view or access your specific account configurations, including your pipelines, configurations, and secrets.
  • No action execution: Cannot create or modify resources. Use the DevOps Agent for that.
  • Public documentation only: Does not have access to internal or private documentation.

For issues not covered in the documentation, go to Harness Support. You can also submit feedback through the UI by selecting Help > Give us feedback.