Skip to main content

Platform Assistant

The Waxell dashboard includes a built-in AI assistant that can query your observability data, create and manage policies, analyze costs, and troubleshoot agent issues -- all through natural language.

Click the Ask AI button in the top navigation bar or the floating chat icon to get started.

What It Can Do​

Query Your Data​

Ask questions about your agents, runs, and costs in plain language:

"How much did my agents cost in the last 24 hours?" "Which agent has the highest error rate this week?" "Show me the slowest runs for support-bot" "What models is my RAG pipeline using?"

The assistant queries your actual observability data and returns real numbers, not generic advice. It can render charts inline for visual analysis.

Create and Manage Policies​

The assistant can create policies interactively:

"Create a rate limit policy for support-bot, max 50 requests per hour" "Add a daily cost budget of $10 for all agents" "Set up business hours restrictions for the enterprise team"

When you ask for a policy, the assistant renders an interactive policy card showing the full configuration. You review and confirm before anything is created -- the assistant never creates policies without your explicit approval.

You can also modify existing policies:

"Increase the token budget for support-bot to 10,000" "Disable the rate limit on my dev agents"

Review Governance​

Get a quick governance health check:

"Run a governance check" "What recommendations do I have?" "Show me all policy violations from today" "Which agents don't have any policies?"

Troubleshoot Issues​

Investigate problems with your agents:

"Why did support-bot fail 3 times today?" "Show me the error traces for my RAG pipeline" "Compare the costs of gpt-4o vs claude-sonnet-4 for my agents"

"Take me to the policy editor" "Where can I see session analytics?" "Open the governance policies page"

Suggested Prompts​

The chat panel shows four suggested prompts to get started:

  • Fleet briefing -- Overview of all agents, recent activity, and health
  • Cost breakdown -- Spending by agent, model, and time period
  • Error analysis -- Recent failures, patterns, and affected agents
  • Governance check -- Policy status, violations, and recommendations

Interactive Policy Cards​

When the assistant proposes a policy, it renders as an interactive card with buttons:

ButtonAction
Create PolicyCreates the policy on the control plane
Open in EditorOpens the full policy editor with the config pre-filled
Suggest AdjustmentsAsks the assistant to modify the policy
Ask About This PolicyGet more context on what the policy does

This makes policy creation conversational -- iterate on the configuration through chat until it's right, then create it with one click.

How It Works​

The assistant is a purpose-built AI agent running on Claude Sonnet. It has access to:

  • Knowledge base -- RAG-powered docs search for platform help
  • Tenant data -- Read-only access to your runs, costs, errors, policies, sessions, users, prompts, and audit logs
  • Policy management -- Validated create/update operations with user confirmation
  • Navigation -- Can link you to the right page in the dashboard

All conversations are scoped to your tenant -- the assistant only sees data you have access to. Conversation history is maintained per-user for context continuity.

API Access​

The assistant is also available via API for programmatic integration:

# Send a message
curl -X POST -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"message": "How much did my agents cost today?"}' \
  "https://acme.waxell.dev/api/v1/assistant/chat/"

# Poll for response
curl -H "Authorization: Bearer $TOKEN" \
  "https://acme.waxell.dev/api/v1/assistant/chat/{signal_id}/result/"

# Get conversation history
curl -H "Authorization: Bearer $TOKEN" \
  "https://acme.waxell.dev/api/v1/assistant/chat/history/"

Next Steps​