Docs Help
JD
New agent
What kind of agent
do you want to build?
Choose a persona to get started. Your Blueprint is pre-configured the moment you pick.
Blueprint · —
Pick a persona to
initialise the blueprint
New agent
What's your
agent's name?
Name your agent. Watch the Blueprint update as you type — this already feels like yours.
The name appears across the platform and in all agent outputs.
📘
The IDENTITY block in the Blueprint updates live as you type. Your agent's name will appear everywhere across the platform.
Blueprint · —
First run
Atlas is working through your queue
Watch your agent process a realistic mock scenario. Each action is annotated with its reasoning.
Blueprint · —
Shift report
Here's what Atlas handled
in the last 24 hours
Your agent's first shift report. This is what it would look like in a real working day.
47
Processed
38
Auto-resolved
9
Escalated
1m 42s
Avg handle
⚡
94% accuracy — Atlas correctly resolved or routed 44 of 47 items without any human guidance on the first attempt.
Based on realistic mock data for your persona — no integration required
Blueprint · —
Rules editor
Atlas built a starter kit
from the First Run
Three rules were auto-discovered from patterns in your mock run. Edit, delete, or add your own — or proceed straight to the test.
Billing queries → priority high
Agent-discovered
Detected during the First Run: billing-tagged tickets consistently needed faster handling.
Enterprise customers → escalate to tier-2
Agent-discovered
Enterprise accounts appeared in 3 of 9 escalation cases. The agent identified the pattern and built this rule automatically.
Untagged tickets → auto-tag + flag
Agent-discovered
22% of incoming items had no tag. The agent now auto-tags based on content and surfaces them for a quick review.
Blueprint · —
Test run
Atlas is running end-to-end scenarios
Simulated test cases based on your persona and the rules in your Blueprint. 5 of 6 passed — you're ready.
Billing dispute — long-term enterprise customer
Pass
Atlas applied billing_priority rule correctly. Priority set to high within 0.3s.
SSO failure — 12 affected users
Pass
Pattern matched identity team routing. Escalated with full context. 100% match.
Untagged inbound — no topic signal
Pass
auto_tag_untagged rule triggered. Content-based tag applied: billing. Flagged for review.
Edge case: enterprise customer with billing issue
Pass
Both billing_priority and escalate_vip fired correctly. Priority: high + escalate tier-2.
Hard case: ambiguous request, low confidence
Fail
Atlas confidence 58% — below threshold. Escalated correctly but missed suggested response template.
Hard case: data export failure at scale
Pass
Data integrity pattern detected. Severity: high escalation triggered in 0.8s. Correct.
✓
All critical cases passed. One non-critical case flagged for review. You can jump back to Rules or proceed to create your agent.
Blueprint · —
Blueprint · —
Project
Overview
Agents
Integrations
Analytics
Account
Settings
Team
Blueprint
Acme Corp / Support Project
🎉 Atlas has been created
Agents
🎫
Atlas
Triage Agent · Created just now
Active
💬 Test your agent
Open a chat session to test how your agent behaves before going live.
🚀 Deploy to environment Later
When you're ready, connect your real systems here and deploy.
Day 3 — Integration nudge
Atlas has been running
for 3 days. Ready to go live?
You've completed 6 test sessions and everything looks solid. Connect your real systems and deploy.
Day 0
Created
Day 1–2
Testing
Day 3
Now
Day 5
Follow-up
In-app banner
🔗
Atlas has been running for 3 days. Ready to go live?
You've completed 6 test sessions. Connect your real systems and deploy to your live queue — Atlas is ready.
Parallel email