Skip to main content

AI settings

Last updated: November 2026

Project Settings > AI controls the project context that Katalon AI uses before it answers or acts. The settings include the Knowledge Base, Agent Rules, Project Memory, and MCP & Skills.

Requirements
  • An account administrator must enable AI in System > Configuration > AI Service.
  • Your role must have the project AI permission to view and use these settings.
  • A project administrator is required to enable Agent Rules and Project Memory.

How the settings work​

  • Knowledge Base provides project data and uploaded documents as reference material.
  • Agent Rules define project-wide standards that Katalon AI must follow. A known issue allows a conflicting personal memory to override a rule.
  • Project Memory stores your personal, durable preferences for the project. Other project members cannot see your memories.
  • MCP & Skills provides the MCP endpoint and setup instructions for connecting external AI clients and Katalon True skills.

Account administrators control which AI features are available to your account, and can view a history of changes to AI settings and keys. See Choose which AI features are available. Changes to project-level settings are recorded in a backend audit stream. There is no project-level audit view.

Knowledge Base​

The Knowledge Base combines project data with documents uploaded to the project. After indexing, Katalon AI Assistant and connected agents can use the content within your existing permissions.

A project's knowledge base gets its information from two sources:

SourceWhat it holdsHow it gets there
Project dataTest cases, test suites, test folders, and requirements already in the projectSynced automatically when ingestion is on
Your documentsFiles you upload, such as PRDs and specsUploaded by you

Both sources are available through the same knowledge base. Katalon AI Assistant and agents connected through the MCP server can access only what you already have permission to access. Turning on the knowledge base does not grant anyone additional access.

Set up the Knowledge Base​

Knowledge Base ingestion is Off by default for new projects. Turn on Enable Knowledge Base before uploading documents.

  1. Go to Project Settings > AI > Knowledge Base.
  2. Switch on Enable Knowledge Base. The page displays the current status, last update time, and counts for synced test cases, test suites, test folders, and requirements.
  3. In Your documents, select Upload files, or drag files onto the upload area.
  4. Wait until each document shows Completed. It is then available to Katalon AI Assistant and connected agents.

After the initial sync, changes to the project are synced automatically about every 15 minutes.

Use this interactive walkthrough to see the complete setup:


note

Turning ingestion off stops future syncing, but does not delete data that is already in the knowledge base.

Manage documents​

  • Supported file types and size: .pdf, .docx, .xlsx, .pptx, .txt, .csv, .md, .png, .jpg, .jpeg. Each file can be up to 100.0 MB.
  • Each filename must be unique within the project. Rename a duplicate file, or use Replace on the existing document's row.
  • Use the Search filename field to find a document. The dropdowns next to it let you filter by file type and status.
  • Ingest status: Documents go through several processing stages. Only documents marked Completed are available to Katalon AI Assistant. If processing fails, hover over the status to see why.
  • Select Clear all to remove all uploaded documents and synced project data from the knowledge base.

Deleting cannot be undone. You must upload your documents again, and project data must go through a new full sync.

Conflict checks​

When Katalon AI Assistant generates test cases or analyzes a requirement, it checks the requirement against the documents and project data in the knowledge base and ends its response with a Potential conflicts section. If the knowledge base is empty, rebuilding, failed, or out of date, the response says it could not check. To learn more, see Review potential conflicts.

Known limitations​

  • If you permanently delete a project item, it may remain in the knowledge base until the next full sync.
  • When generating test cases, the knowledge base does not yet help prevent duplicate test cases.

Agent Rules​

Agent Rules set project-wide standards for Katalon AI. Turn on Enable Agent Rules to apply them. You can create Agent Rules only on this page. Asking Katalon AI to remember something in chat creates a personal memory in Project Memory.

  • To add a rule, select Add the first rule or Add rule. Enter a required Rule, optional Details, and a Tag. Use Details to explain when the rule applies; vague notes such as "Added for testing" can make it unclear to Katalon AI.
  • To manage rules, use the search field, Edit, Delete, or Export.
  • To import rules, select Import and upload a CSV file with Rule and Tag columns. Details is optional.

Katalon AI identifies the rules it applied in its response.

Note: A conflicting personal memory can override an Agent Rule due to a known issue.

Rule tags​

The tag tells Katalon AI what kind of statement the rule is. It does not change how strictly Katalon AI follows the rule.

TagUse it forExample rule
conventionHow your team does the work: naming, structure, format, where things go. If Katalon AI ignores it, the result works but does not match your standard.Test case titles use the format <Feature> - <Scenario>.
constraintA hard limit. If Katalon AI ignores it, the result is wrong, unsafe, or unusable.Never run tests against the production environment.
environmentA fact about the application or system under test. It describes the system and gives no instruction.The staging database resets every day at 02:00 UTC.
correctionA mistake Katalon AI made before, with the right answer.The checkout page in this app is called "Basket". Do not call it "Cart".
verdictHow to judge a type of result, so Katalon AI decides the same way each time.Classify a timeout on the payment step in staging as an environment issue.
preferenceA choice with no team standard behind it. For a personal choice, use Project Memory instead.Keep generated test cases under 10 steps.

Rule of thumbs​

  • If the rule states a fact about the system under test, select environment.
  • If ignoring the rule makes the result wrong or unsafe, select constraint.
  • If the rule fixes a mistake Katalon AI made before, select correction.
  • If the rule tells Katalon AI how to judge a result, select verdict.
  • If the rule describes how your team names or structures work, select convention.
  • If you are not sure, select convention. The tag does not change how the rule applies.

A rule about AI test runs can mix a fact and an instruction. Split it into separate rules. For example, "The QA environment uses Okta single sign-on" is an environment rule, and "Do not run AI tests on the shared QA environment during business hours" is a constraint rule.

Sample prompts​

Use these prompts in Katalon AI Assistant to see how a rule of each tag affects the result. Each prompt assumes the example rule from the table above is active, with details that explain it. Katalon AI applies rules in chat, including when it sets up a Run with AI for you.

Show sample prompts
TagPromptExpected result
conventionWrite test cases for the login page.Each test case title uses the format <Feature> - <Scenario>, and Katalon AI lists the rule under Rules applied.
constraintRun the checkout suite with AI on production.Katalon AI does not start the run and tells you which rule stops it.
environmentWhen should we schedule the nightly regression run on staging?Katalon AI suggests a time after the 02:00 UTC database reset.
correctionWrite test cases for the cart page.The test case titles and steps use "Basket".
verdictA staging run failed with a timeout on the payment step. Should I file a defect?Katalon AI classifies the failure as an environment issue and says that no defect is needed.
preferenceGenerate test cases for the password reset flow.Each test case has fewer than 10 steps.

When Katalon AI generates test cases, its reply lists the rules it followed under Rules applied. To see all rules in force, ask "Which agent rules apply to this project?"

Note: If a new rule has the same details text as an existing rule, the new rule is not saved, even if the import reports it as added.

Project Memory​

Project Memory stores your own durable preferences for the project. Turn on Enable Project Memory, then ask Katalon AI to remember a preference, such as β€œRemember that our checkout tests use the staging environment.” You can also manage memories from ... > Project memory in the Katalon AI header.

Project memories are personal and are not visible to other project members. A conflicting personal memory can override an Agent Rule due to a known issue.

MCP & Skills​

Your MCP server​

The Your MCP server section provides one endpoint for the whole account. The same endpoint works for every project in the account. Select Copy to copy the endpoint. On first use, your agent signs you in through the browser and asks you to choose a workspace.

Set up your agent​

Setup varies by client. Select your client to view its setup command or configuration, then select Copy and run it in the client. The available clients are Claude Code, GitHub Copilot, Codex, Cursor, Claude Desktop, and ChatGPT. For Claude Code and Codex, the skills and MCP server are installed together.

After setup, the client signs you in through the browser on first use and asks which workspace to use.

Katalon True skills​

The page links to the Katalon True skills repository. It provides 18 reusable skills covering the testing lifecycle, including planning, design, management, review, execution, analysis, and maintenance. The skills help agents use your test assets, requirements, and runs according to your project conventions.

Was this page helpful?