Branching DAG Engine
Topological execution with `if`, `switch`, and `try/catch` for controlled, readable AI logic.
Visual AI Automation
Stop stitching tools together. Build one reliable AI flow.
L2M helps teams design, execute, monitor, and automate production-ready AI workflows with deterministic branching, guardrails, human approvals, and full execution traceability.
Capabilities
Everything your orchestration layer should have
Execution History
Persistent audit logs with node-by-node trace for debugging, compliance, and quality review.
Human-in-the-Loop
Pause on approval nodes, serialize state, and resume safely after approve or reject decisions.
Guardrails
Validate input and sanitize model output with deterministic checks and configurable failure behavior.
Streaming Chat + Widget
SSE-powered chat interface in-studio and an embeddable widget for external websites.
Code Node Sandbox
Run custom JavaScript transforms in a VM-isolated execution step with timeout controls.
Batch 5
Native cron scheduling for autonomous workflows
Add a `schedule_trigger` node to run workflows every day, every 15 minutes, or any cron interval with timezone support.
- Scheduler initializes active workflows on API startup
- Each cron run executes with `triggerType: "cron"` and is saved in execution history
- Schedules reload dynamically when workflows are saved, imported, or deleted
Common Schedules
0 9 * * *
Daily standup summary
*/15 * * * *
Quarter-hour monitoring loop
0 0 * * 1
Weekly Monday report
Stack
Practical architecture for fast iteration
FrontendVite + React Studio
BackendFastify + SQLite
EngineWorkflow DAG Executor
IntegrationsRAG, MCP, LLM Providers
From local prototype to production automation
L2M is open-source and built for teams that want transparent AI workflow logic instead of opaque chains.