Powered by Gemini 3 Flash

Design AI Agents Visually.
Deploy with Confidence.

Translate plain English prompts into visual graphs to truly understand what your prompt is doing. Build single agents or full multi-agent systems with master orchestrators and linked sub-agents. Catch hidden errors, empty paths, warnings, and dangerous actions before you deploy.

22
Node Types
5
Templates
5
AI Providers

Workflow

How It Works

From a plain English description to an analyzed, executable agent graph — in four steps.

1
Describe

Type a natural language prompt describing the logic you want to visualize. Translating it to a graph helps you truly understand what your prompt is doing — catching empty paths, warnings, and dangerous actions.

"Create a graph for a support agent that routes billing to humans, handles tech issues automatically... to help me understand the logic and catch errors."
2
Generate

Gemini 3 Flash reads your prompt and builds a complete node graph — typed, Dagre-positioned, and connected.

12 nodes created & typed
9 edges connected
Conflicts flagged in real time
3
Edit & Chat

Drag nodes visually or type commands in the chat panel — "Add a retry loop around Search Tool".

Visual + conversational editing
4
Analyze & Walk Through

AI analyzes conflicts, auto-fixes issues, then simulates your agent in Simulation Studio — two modes, data flow tracking, and rich error reporting.

AI-narrated logic walkthrough

Visual Graph Editor

Your agents, laid out visually

The interactive canvas uses intelligent auto-layout via Dagre — nodes are automatically positioned in a clean top-down flow. Annotation nodes (guards, rules, tools, memory) are styled distinctly with dashed borders. Drag nodes to reposition, connect them with typed edges, and use the minimap to navigate large graphs.

🚀Start
START
👤Customer Persona
PERSONA
🛡Safety Guard
GUARD
Classify Intent
TASK
📋Billing Rules
RULE
Intent Decision
DECISION
🔧Search Tool
TOOL
Human Handoff
HANDOFF
🎬Billing Lookup
ACTION
No conflicts

Prompt Architecture

Visualize Your Prompt Logic

Paste your agent's system prompt or natural language logic. Verto maps the instructions into a structural graph, allowing prompt engineers to audit how the AI interprets boundaries, tools, and transitions. See the "Mental Model" of your agent instantly.

Analyze Agent Prompt
"You are a Support Agent. Primary goal: Resolve billing issues. If user asks about tech, use TECH_TOOL. If issue unresolved > 3 turns, ESCALATE. Maintain safe tone at all times..."
Analyzing logic structure with Gemini...

Structural Audit → Mental Model

Verto translates your text instructions into a verifiable architecture. Audit the branching logic, verify safety guards are in-path, and ensure no orphaned edge cases.

🤖
Logic Mapped
Full hierarchical flow of instructions
🔗
Context Preserved
Original prompt bound to graph nodes
🔄
Quantifiable Drift
Track consistency between text & graph

LLM Output Formats

Three Formats. One Graph.

Verto supports three output formats from the LLM — each producing the same graph but optimized for different trade-offs between readability, token efficiency, and generation speed.

JSON
Standard

Full verbose JSON with all fields explicitly spelled out. Maximum readability, ideal for debugging and manual inspection.

{
"metadata": {
"agent_id": "support_bot",
"persona": "Support Agent"
},
"graph": {
"nodes": [...],
"edges": [...]
}
}
~6K chars~2K tokens
YAML
Human-Friendly

Indentation-based format with no braces or brackets. Excellent readability, slightly more tokens than JSON.

metadata:
agent_id: support_bot
persona: Support Agent
graph:
nodes:
- id: n1
type: start
label: Start Agent
edges:
- source: n1
~8K chars~2.2K tokens
JSON Compact
Fastest

Array tuples with type codes and integer IDs. Decoder expands to full graph. Up to 66% fewer tokens.

{"m":[...],"g":{
"n":[
[1,"st","Start","Entry","c"],
[2,"d","Route?","Check type","c"],
...
],"e":[
[1,2,"Next"],
...
]}}
~2.4K chars~780 tokens
Benchmark Comparisongemini-3-flash-preview
JSON
6,106 chars1,931 tok
YAML
8,102 chars2,265 tok
JSON Compact
2,407 chars778 tok
JSON Compact: 61% fewer chars, 60% fewer tokens vs JSON

Graph Chat

Edit Your Graph by Chatting

The Graph Chat panel sends your instruction plus the full graph context to Gemini. It interprets commands like "add a retry loop around Search Tool" and generates precise graph mutations — adding, removing, or updating nodes and edges. Every edit shows stats: nodes added, removed, or updated.

Graph Chat
AI-Powered
Add a data validation step after the intake node
Verto
Added "Validate Input Data" (GUARD) after "User Message" with a connection to "Classify Intent". The node checks for required fields and format compliance.
+2 nodes+1 edges
Add a retry loop around the Search Tool with max 3 attempts
Verto
Added a CONDITION node before "Search Tool" with a retry counter in CONFIG. After 3 failed attempts, the CONDITION routes to a HANDOFF node for human escalation.
+2 nodes+1 edges
Type a command to edit the graph…

Ctrl+Enter to send

Conversational Graph Editing

The Graph Chat panel uses Gemini to understand your instructions in context. It reads your entire graph, figures out the right mutation, and applies it — adding, removing, or updating nodes and edges.

Add nodes
"Add a data validation step after the intake node"
Modify nodes
"Change the retry count on Search Tool to 5"
Remove nodes
"Remove the logging node, it's not needed"
Restructure
"Wrap the tool calls in a retry loop"

Bidirectional Re-sync

Compare Any Two Versions

Re-sync reconstructs a prompt from any graph state — no AI call needed. Use the dropdowns to pick which two sources to compare: your original prompt, any saved version snapshot, or the live graph. Word-level diffs and a similarity score quantify exactly how much things have drifted.

Prompt Re-sync
87% match
2 added1 removed
Original Prompt
1You are a customer support agent.
2Classify incoming user intents accurately.
3Route billing queries to human agents.
4- Store interaction history to memory after each turn.
 - Applies safety filter before all responses.
 - Blocks inappropriate content automatically.
5Escalate unresolved cases after 3 failed attempts.
v2.2 — Updated safety thresholds
1You are a customer support agent.
2Classify incoming user intents accurately.
3Route billing queries to support agents.
 - Store interaction history to memory.
4- Applies safety filter before all responses.
5- Blocks inappropriate content automatically.
6Escalate unresolved cases after 3 failed attempts.

What Re-sync Does

Use the dropdowns to choose any two sources to diff: original prompt, any saved version snapshot, or the live graph. Re-sync walks the graph using stored source positions, node order, and logic snippets to reconstruct a faithful prompt — no AI call. The Jaccard word-overlap score gives you a single number: how far apart the two sources are.

Red rows
Lines present in the left source but absent from the right — nodes or rules that were removed or changed between versions.
Green rows
New behaviors in the right-side source that aren't in the left — nodes or rules that were added in this version.
Highlighted words
Individual tokens that changed between original and reconstructed — word-level precision for subtle drift.
Why it matters

Visual graph edits are intuitive but can silently lose intent. Compare any two snapshots — including branched versions — to quantify exactly what changed between any two points in your agent's history.

Branch Version Control

Git-Like Branching History

Every node add, delete, and edit auto-commits a snapshot. Versions branch like git — edits on v1 create v1.1, v1.2; a fresh start creates v2. The SVG branch tree shows exactly which version each commit descended from.

Branching Lineage

Unlike a flat version list, Verto tracks which snapshot each commit descended from. If you restore v1 and make edits, those edits branch off v1 — not from the latest. The visual tree makes this lineage immediately visible.

Auto-commit on every node add / delete / edit
AI generation creates no version — first edit creates v1
Edits on sub-versions create v1.1, v1.2, v1.3…
New branch from original creates v2, v3…
Visual SVG dot-and-line branch tree
Instant rollback to any snapshot
Diff stats: nodes added, removed, modified
Version Control
6 Commits
v2.3
Describe your changes…
Commit
v1Added safety guard and personaYesterday
Restore
v1.1Tweaked intent classifier output format8h ago
Restore
v1.2Removed unused logging node7h ago
Restore
v2Restored v1, added billing escalation4h ago
Restore
v2.1Retry loop around Search Tool2h ago
Restore
v2.2Updated safety thresholdsJust nowactive

AI Conflict Detection + DAG Validation

Catch Issues Before They Ship

Two validation layers protect your graph. The DAG Validation Engine runs 15 deterministic structural rules instantly — checking for cycles, self-loops, disconnected components, missing source/sink nodes, and topological sortability. The AI Conflict Analyzer uses Gemini for LLM-powered risk detection, guard bypass analysis, and semantic reasoning — finding unguarded dangerous actions, logic contradictions, range gaps, and prompt/graph divergence. Click any detected action to navigate directly to its node on the canvas. Filter by severity and auto-fix issues with one click.

AI Conflict Analyzer
1 Critical1 Warning1 Info
Prompt QualitySafety & Permissions1Structure
Filter:critical (1)warning (2)
Actions & Permissions3 actions
Search External APIExternal APIGuarded
Delete User DataData MutationUnguarded
Send NotificationUser CommsGuard bypassed
1 unguarded high-risk action · 1 guard bypass detected
Guard can be bypassed: Send Notificationcritical

"Send Notification" has a GUARD node, but there exists an alternative path that reaches this action without passing through the guard.

3 actions · 1 issue · Click any action to navigate to its node
Fix all auto-fixable
Critical
  • No self-loops
  • No directed cycles
  • Source node exists
  • Sink node exists
Warning
  • Topological sortability
  • Reachability from sources
  • Path to sink exists
  • No disconnected components
  • No duplicate edges
Info
  • Handshaking lemma check
  • High-degree node alerts
  • Transitive reduction hints

What It Detects

15 deterministic DAG rules run instantly. The AI analyzer uses Gemini to detect risky actions, classify their permissions, and find guard bypass paths — plus semantic reasoning for issues simple linting would miss. Click any permission to pan directly to its node. Enable DAG-aware AI generation in Settings > Graph Rules to have Gemini actively avoid violations during creation and editing.

LLM-powered risk & permission detection
Guard bypass path analysis
Unguarded high-risk action warnings
Missing escalation paths
Numerical range gaps & overlaps
Orphaned / unreachable nodes
Circular dependencies & self-loops
Stale thresholds & outdated values
Contradictions with original prompt
Missing start / end / source / sink nodes
DAG cycle risks & implicit loops
Semantic reachability concerns
Disconnected graph components
Settings > Graph Rules
Post-parse validation

Validate immediately after AI generation

ON
Strict chat edit mode

Reject edits that create critical DAG violations

OFF
Pre-flight runner check

Validate before executing an agent

OFF
DAG-aware AI generation

Inject DAG rules into AI prompts (~1.5-2x tokens)

OFF
Auto-fix

For fixable issues, the AI generates a precise graph mutation — adding missing edges, inserting guard nodes, or updating stale values — applied with one click.

Simulation Studio

Test Your Agent Before You Ship

A unified simulation environment with two modes, pre-flight validation, data flow tracking, and rich error reporting — all in a 3-column layout.

Simulation Studio
Test your agent with sample inputs and visualize execution flow
Preview
LLM Simulation
Each executable node = 1 LLM call. Uses more tokens than Preview.
Test Input
My invoice shows the wrong amount for February
Generate
Save
Controls
Run
Saved Test Cases
No saved test cases yet
Pre-flight: Structure OK 1 warning|Runtime: 1 warning
Step Timeline
5 step(s)
1.Start
START
2.Safety Guard
GUARD
Took: passed
3.Classify Intent
TASK
4.Intent Decision
DECISION
Took: Billing Flow
5.Billing Lookup
ACTION
No outgoing connection
Data Inspector
TASKClassify Intent
Input
"My invoice shows the wrong amount for February"
Output
{ "intent": "billing_dispute", "confidence": 0.94 }
Data Changes2
+intent:"billing_dispute"
+confidence:0.94
Conditions3
billing_dispute
general_inquiry
cancellation

Preview Mode

1 LLM call total

Makes a single Gemini call to generate realistic sample data for every node in your graph, then walks through them instantly. Fast, cheap, and great for validating structure.

SpeedInstant walkthrough after initial generation
CostLow — one API call regardless of graph size
Best forQuick structure checks, path validation, cycle detection

LLM Simulation

1 LLM call per executable node

Runs your agent like real execution — each TASK, ACTION, DECISION, and GUARD node gets its own Gemini call. Produces realistic per-node outputs and real decision evaluations.

SpeedReal-time, step by step with streaming
CostHigher — token usage scales with node count
Best forEnd-to-end testing, realistic output validation
Pre-flight checks
Validates graph structure before running
Data flow tracking
See what changed at each step
Path decisions
Track which branches were taken and why
Error reporting
Cause, message, and fix suggestion

Comprehensive Taxonomy

22 Node Types for Every Scenario

Every part of an AI agent workflow has a purpose-built node — from personas and triggers to memory stores, safety guards, and logging outputs.

🤖
AGENT
Autonomous LLM agent
📋
RULE
Business logic gate
TASK
Unit of work
HANDOFF
Agent-to-agent pass
🔧
TOOL
External API call
💾
MEMORY
State persistence
🛡
GUARD
Safety filter
TRIGGER
Event activation
🔀
CONDITION
Branch condition
🎯
RESOLUTION
Outcome handler
🚀
START
Flow entry point
👤
PERSONA
Agent identity
⚙️
CONFIG
Runtime settings
DECISION
Routing decision
OPTION
Choice branch
👣
STEP
Sequential action
📚
REFERENCE
External pointer
🎬
ACTION
Execute action
END
Flow termination
📥
INPUT
User / data input
🗃️
LOGGING
Observability output
📂
GROUP
Container for grouped nodes

Quick Start

Start From a Template

Five pre-built agent templates get you to a working graph instantly — no prompt needed.

🎧
Customer Support
Intent classification, routing, and escalation for support workflows.
AGENTTASKRULEHANDOFF
5 nodes · ready to use
Approval Workflow
Request validation, auto-approval, manager review, and notifications.
DECISIONGUARDSTEPACTION
5 nodes · ready to use
🛡️
Content Moderation
Profanity filter, spam detection, sentiment analysis, and publication gate.
GUARDRULECONDITIONACTION
5 nodes · ready to use
⚙️
Data Pipeline
Extract, validate schema, transform, and store to memory.
TASKTOOLMEMORYSTEP
5 nodes · ready to use
🔗
Multi-Agent
Master orchestrator delegates to specialist sub-agents, each with its own graph — click through to drill into any sub-agent.
AGENTHANDOFFTRIGGERRESOLUTION
5 nodes · ready to use

Full Feature Set

Everything You Need

From first draft to production-ready agent — Verto covers the entire workflow.

AI Agent Generation
Type a description and Gemini 3 Flash builds a complete, structured agent graph in seconds. Optional DAG-aware mode injects structural rules so the AI avoids cycles and disconnected graphs.
Graph Chat Editor
Edit your graph by chatting — "Add a validation step after intake" — and watch nodes appear instantly.
Bidirectional Re-sync
Reconstruct a prompt from the visual graph. Compare any two sources — original prompt, a saved version, or live graph — with word-level inline diffs and a similarity score.
Visual Graph Editor
Interactive drag-and-drop canvas with zoom, pan, minimap, and intelligent auto-layout via Dagre. Supports GROUP container nodes for hierarchy.
DAG Validation Engine
15 formal DAG rules: acyclicity, topological sort, self-loops, source/sink, reachability, disconnected components, handshaking lemma, transitive reduction hints, and more.
Branch Version Control
Git-like branching history with auto-commits on every edit. Visual SVG branch tree shows parent-child lineage (v1 → v1.1 → v1.2, or branch to v2). Rollback to any snapshot in one click.
Complexity Metrics
Cyclomatic complexity, branching factor, max depth, and actionable optimization suggestions.
Node Comments
Add annotations to individual nodes — document intent, flag edge cases, and keep design notes alongside the graph.
Export & Import
Export as JSON, import from file, paste raw JSON, and toggle between visual graph and text editor.
Undo / Redo & Auto Layout
Full command history with Ctrl+Z/Y, plus hierarchical and force-directed graph auto-layout.
Keyboard Shortcuts
Full shortcut reference: Cmd+S save, Cmd+Z/Y undo/redo, Cmd+C/V/D copy/paste/duplicate, Delete to remove, and ? to open the dialog.
Multi-Agent Systems
Build master-orchestrator + sub-agent architectures. Auto-detect multi-agent prompts, generate linked graphs, and drill into any sub-agent from the master canvas with breadcrumb navigation.

Multi-Provider AI

Works With Your AI Stack

Plug in your own API key for any supported provider. Default is Gemini 3 Flash.

Gemini 3 FlashDefault
OpenAIGPT-4o & more
AnthropicClaude models
GroqUltra-fast inference
Custom EndpointAny OpenAI-compatible API

In Development

Coming Soon

Features currently in development. Expect rough edges and breaking changes.

Simulation StudioWIP
Unified 3-column simulation environment with two modes: instant preview (1 LLM call) and full step-by-step LLM execution. Data flow diffs, condition tracking, pre-flight validation, and rich error reporting.
AI Conflict AnalyzerWIP
LLM-powered risk detection with guard bypass analysis, severity filters, click-to-navigate permissions, and one-click auto-fix. Detects range gaps, contradictions, unguarded actions, and structural issues.
Agent HubWIP
Share agents publicly, discover community builds, star ratings, and search by capability.
Pattern LibraryWIP
Insertable sub-graph patterns — Chain of Thought, Self-Critique Loop, Input Guard, and more. Browse by category, search by tag, and drop directly onto any canvas.
Experimental Chat EditorWIP
Next-gen AI graph editing with deeper context awareness and multi-step reasoning. Highly experimental — expect rough edges.
MCP ServerWIP
A standalone MCP server exposes all Verto tools — create, edit, validate, run, and analyze agents — accessible from Claude Desktop, Cursor, or any MCP-compatible client.
🚀

Generate a graph from your prompt in minutes.

Translate your prompt into a graph to truly understand what it's doing. Expose hidden errors, catch dangerous actions, and identify empty execution paths before they break your application.

Open Agent Architect →

No account required · Runs entirely in your browser · Free to use

Have Questions or Feature Ideas?

I'm constantly improving Verto Agent Architect. If you encounter any issues, have suggestions for new features, or just want to chat about agent design, feel free to reach out.

project.verto@proton.me