Tool Module

Diagrams Tool

Generate diagrams from natural language.
Visual communication for AI agents.

Inactive
January 2026
The Vision

Agents describe,
Kroki renders

Give AI agents the ability to create architecture diagrams, flowcharts, and visualizations by simply describing what they want to show.

Versatility

8 diagram types supported

Flowchart
Process flows and decision trees
Sequence
Component interactions over time
Class
Object-oriented design structures
ERD
Database entity relationships
Architecture
C4 model system diagrams
Mindmap
Hierarchical concept maps
Gantt
Project timelines and schedules
State
State machines and transitions
Defined via DiagramType enum across 8 variants
Architecture

Zero dependencies

Kroki.io API
No GraphViz installation. No PlantUML server. No local rendering engine. Just HTTP requests to Kroki's free public API.
Self-Hosted Ready
Point to your own Kroki instance for air-gapped environments or enterprise deployments. Same API, full control.
Agents generate diagram definitions in Mermaid, PlantUML, or GraphViz syntax. Kroki handles rendering to SVG/PNG. Simple, decoupled, reliable.
Usage

Generate a diagram

# Agent describes what to show { "name": "generate_diagram", "parameters": { "diagram_type": "sequence", "description": "User login flow with JWT authentication", "format": "svg" } } # Tool generates Mermaid definition sequenceDiagram User->>Frontend: Enter credentials Frontend->>Backend: POST /login Backend->>Database: Verify user Database-->>Backend: User valid Backend->>Backend: Generate JWT Backend-->>Frontend: Return token Frontend-->>User: Login success # Kroki renders to SVG
Implementation

Three-component design

Kroki Client
HTTP wrapper handling API calls, response parsing, and format negotiation. 265 lines in kroki_client.py.
Template Loader
YAML-based template management with validation and variable substitution. 168 lines in template_loader.py.
Validation
Syntax checking before rendering, type enforcement via models.py, and diagram definition validation.
Modular design allows easy extension. Adding new diagram types or templates is just a YAML file. The core diagram engine is 320 lines in diagram.py.
Development Snapshot

A focused prototype

1
Commit
1,650
Lines of Python
35
Test Functions
1
Contributor
12 Python files across source and tests
Single commit 923ca3b on January 21, 2026
Potential Adoption

Bundles that could adopt diagrams

designer
Generate wireframes, user flows, and system architecture diagrams during design phases
full-stack-developer
Document APIs, database schemas, and component interactions automatically
content-creator
Create infographics, process diagrams, and visual explainers for documentation
spec-driven-builder
Generate architecture diagrams from specifications before writing code
Note: Bundle adoption claims are unverified against live bundle configurations.
Sources & Methodology

Research transparency

Data as of
February 20, 2026
Feature status
Inactive — repo located in ~/dev/ANext/Inactive/
Git log analysis
git log --oneline on amplifier-module-tool-diagrams → 1 commit found (923ca3b)
Line counts
find . -name "*.py" | xargs wc -l → 1,650 lines across 12 files
Test function count
grep -r "def test_" tests/ --count → 35 test functions across 4 files
Contributor analysis
git log --format="%an" | sort | uniq -c → 1 contributor (Sam Schillace, 100%)
Gaps: Test pass/fail rates not verified (pytest not run). Bundle adoption claims not verified against live bundle configs.
Primary contributor: Sam Schillace (1 commit, 100%)

Visual communication
for AI agents.

8 diagram types. Zero dependencies.
Agents that can show, not just tell.

amplifier-tool-diagrams
Part of the Amplifier ecosystem
More Amplifier Stories