★Start Free★
Guides

Using Agentic Chat

Suggest Edits

Agentic Chat is the interactive interface where you can ask questions, explore your environment, generate dashboards, and create agents from natural-language requests.
Sera uses Jit’s knowledge graph and your connected integrations to produce contextual, structured, and actionable output.

Agentic Chat is ideal for:

  • On-demand analysis
  • Exploratory questions
  • Building dashboards
  • Investigating security findings
  • Prototyping an agent before saving it

4.1 Asking Questions

You can type any security-related question into the chat input.
Sera supports natural-language queries such as:

  • “Run a full risk assessment.”
  • “Which apps are most exposed right now?”
  • “Show me misconfigurations across AWS and GitHub.”
  • “Create a dashboard of SLA-violating findings.”
  • “Summarize our supply-chain risk.”

Sera interprets your intent, retrieves the relevant data, and generates structured results based on the context.

Using Tools from Chat:
Certain prompts in Chat may trigger built-in tools automatically when the task requires them.
For example:

  • Jira tools may be used to fetch or create tickets
  • AWS queries may run when asking about cloud resources
  • Slack messages may be sent for urgent alerts

Tools are not configured separately in Chat.
They are invoked implicitly when your question requires a real action (fetching data, creating work, or alerting a team).

For details on each tool’s capabilities and limitations, see Tools.


4.2 Suggested Prompts

Agentic Chat includes suggested prompts underneath the input field to help you get started quickly.

Examples include:

  • Run a full risk assessment
  • Which apps are most critical and exposed?
  • What’s our OSS supply-chain risk?
  • How vulnerable are we to cloud misconfigs?

These prompts demonstrate the types of questions the agent can answer, and can be used as templates for your own queries.

4.3 Saving a Chat as an Agent

Any conversation can be turned into a reusable agent by selecting Save Agent.

Saving a chat:

  • Extracts the underlying task from your request
  • Prepares a reusable workflow
  • Opens the agent configuration modal
  • Allows you to adjust schedule, notifications, Slack output, and advanced settings

For more information on configuration fields, see Section 2.2 — Agent Configuration.

This is the recommended path when you prototype a workflow in chat and want to automate it.

4.4 Sharing a Chat

You can share the results of a chat using the Share button.

Sharing options include:

  • Copy Link — Generates a read-only URL for the conversation
  • Download Run Report — Provides a downloadable summary of Sera’s output

Sharing is useful for:

  • Collaborating with teammates
  • Escalating issues
  • Demonstrating findings
  • Preserving analysis for audit or review

4.6 Note on Accuracy (Beta)

Agentic Chat is currently in Beta.
As with all LLM systems, answers may occasionally be partial, imprecise, or slightly off.
For best results, ask clear and scoped questions.

4.7 What Chat Works Well For

Agentic Chat performs best with prompts that follow predictable, structured patterns—especially those that leverage Jit’s knowledge graph.

Below are the most reliable prompt types, each with examples taken from real-world usage.

1. Relationship Queries

Questions about how assets connect to each other.

Examples:

  • “Which teams own which services and repositories?”
  • “Show me the dependency relationships between our microservices.”

2. Prioritization Queries

Ranking or sorting by risk, severity, or impact.

Examples:

  • “List the top 3 prioritized production GitHub repositories.”
  • “Show me the top 5 high-priority findings in production.”
  • “Show a prioritized list of SCA vulnerabilities.”

3. Filtered / Scoped Queries

Focusing on a specific environment, severity, or condition.

Examples:

  • “Show critical findings that violated their SLA of 90 days as of July 1, 2025.”
  • “Which production repositories have no open findings?”
  • “How many findings are both in production and externally accessible?”

4. Inventory & Count Queries

Listing or counting assets.

Examples:

  • “How many production repositories do we have?”
  • “Show me all repositories.”
  • “List the top 25 GCP accounts with production resources.”

5. Time-Based Queries

Limited to a time window.

Examples:

  • “Detect new S3 buckets created this week.”
  • “What deployment activities happened in the last 7 days?”

6. Context Lookups

Asking what Jit already knows about an entity.

Examples:

  • “What do you know about a repository called ‘smart-payment-service’?”
  • “What security vulnerabilities exist in our critical production services?”

7. Reasoning / Explanation Queries

Asking the model to interpret or analyze known data.

Examples:

  • “Show me 5 FPs detected by DAST on application X and explain why these are FPs.”
  • “What are the 10 most urgent security issues we need to address?”

4.8 Examples of Good Prompts (Quick Reference)

  • “List the top 3 prioritized production GitHub repositories.”
  • “Group the top DAST vulnerabilities by application.”
  • “Detect new S3 buckets created this week.”
  • “How many high-priority findings (score > 60) exist in production?”
  • “What do you know about the repository ‘smart-payment-service’?”

Summary

Agentic Chat is your interactive entry point for exploring Jit, generating insights, and building dashboards.
From here, you can:

  • Ask natural-language questions
  • View structured widgets
  • Prototype complex analyses
  • Save your work as automated agents
  • Share outputs with your team

Sera combines conversational ease with powerful analysis capabilities to help streamline your security workflows.

Updated 24 days ago