Graphs you saved from the canvas (via ⋯ → Save as template)
Built-in templates
4 agents active · idle
6 nodes
5 edges
guard idle
Scroll to zoom · Double-click canvas to add node⌘K commandsv3.0.0
Duplicate
Rename
Delete
Axiom Runtime Guard
Policy packs
Loop governor
Escalation (optional)
The guard is complete offline — it runs the same classifier the hosted firewall runs.
Escalation adds an Axiom-signed manifest for borderline verdicts, nothing more.
Recent manifests
Model registry
Add your own LLM endpoints. API keys are stored locally in your browser.
Add new model
Display name
Provider
Base URL / Endpoint
Model ID
API key
Default temp
Max tokens
Rate in — $ / 1M tokens
Rate out — $ / 1M tokens
Used to price runs. Left blank, runs still report tokens — only the dollar figure is withheld. Provider prices change; these are yours to keep current.
Require API key
Open project…
New graph
Save as template
Import JSON
Export JSON
Export .loop.yaml (git-native)
Copy share link
Welcome & about
Keyboard shortcuts
NodeXLoop
Debug AI agents like you debug code — build, run and govern agent loops without leaving your editor.
Visualize execution live
Set agent breakpoints
Control runtime permissions
Replay & compare every run
Your Code → Agent Loop → Runtime Guard → Verified Result
Paused
Topic input
Research agent
Coder agent
Writer agent
Critic agent
Reader agent
Loop controller
Gate / branch
Merge / join
Final output
Files
Drop files here or click to browse
Text files under 300 KB are read as run context · larger files register as metadata only
Tip: you can also drop files straight onto the canvas — an input node is created automatically. Attach files to nodes from the inspector.
Integrations
Google
Drive & Docs context for research nodes
GitHub
Verified live via api.github.com — repo context for coder nodes
Slack
Post run summaries to a channel via incoming webhook