The Cost Assistant is a chat workspace where you ask questions about your own AWS spending and Parsivex scan results. Answers are grounded in your workspace's latest completed scan — not generic cloud advice or outside knowledge.
Open the full-page workspace from Cost Assistant in the sidebar (/chat), or use the quick Ask about your bill panel in the app header.
Full-page layout
On /chat, the workspace uses a two-column layout:
- Left — past conversations (when persistence is enabled). Click one to reopen it, or New to start fresh.
- Right — the chat panel with suggested prompts, message history, and the input.
Suggested prompts appear when you start a new conversation, for example Which findings should I fix first? and Why did EC2 spike last week?
From the Findings inbox, expand a finding and choose Ask about this finding to open chat with that finding pre-filled in the input (paid plans).
What it can help with
The assistant is designed for questions about your AWS costs and findings, including:
- Spend breakdown — which services drive your bill, how costs changed, and where waste shows up
- Findings — what a finding means, why it was flagged, and how severe it is
- Prioritization — which issues to fix first based on estimated savings in your scan
- Savings estimates — explaining figures from your report (it cites numbers from your scan context only)
- Parsivex concepts — how scans, reports, and realized savings work in your workspace
The assistant only uses data from your latest completed scan. If something is not in that context, it will say so and suggest running a new scan to refresh the data.
What it cannot do
The Cost Assistant is not a general-purpose AI chatbot. It will politely refuse requests outside its scope, including:
| Request type | What to use instead |
|---|---|
| General coding, homework, creative writing, or trivia | Other tools — not Parsivex chat |
| Google Cloud, Azure, or other non-AWS platforms | Parsivex only analyzes AWS accounts you connect |
| AWS CLI, Terraform, or shell commands | Open the finding in your scan report and use its remediation script (paid plans) — see Remediation scripts overview |
| Commands to run in your terminal | Remediation scripts provide reviewed, copy-paste commands with safety warnings |
Refusals are intentional guardrails. They keep answers focused on your bill and prevent unsafe or off-topic output.
Run a scan first
You need at least one completed scan in the workspace before the assistant can answer cost questions. If you open chat before scanning, you will see:
Run a scan first so I can answer questions about your AWS costs and findings.
Run a scan from the dashboard or Run scan in the header, wait for it to finish, then return to chat.
Daily message limits
Chat is available on Monthly Monitoring and Team Plan (chat.cost_assistant entitlement). Each account has a daily cap on billable messages:
| Plan | Chat messages per day |
|---|---|
| Free | 0 (upgrade required) |
| Monthly Monitoring | 50 |
| Team Plan | 200 |
The limit resets at midnight UTC. Your remaining allowance appears above the chat input. When you hit the limit, you will see a rate-limit message — distinct from scope refusals or API errors.
See Plan comparison for the full feature and limit matrix.
Tips for better answers
- Ask about specific services, findings, or savings from your latest report.
- Use suggested prompts when you are not sure where to start.
- Run a new scan after major infrastructure changes so context stays current.