Evaligo

About Evaligo

Evaligo is a web-based visual builder for AI workflows. Users describe or drag-and-drop a flow of prompts, tools and agents as connected nodes, then run the same flow as multiple variants across different models — OpenAI, Google, Anthropic, and open-source models such as DeepSeek, MiniMax, Kimi and Qwen accessed through a single OpenRouter key. Each variant runs on the same inputs, with only the model, prompt or reasoning setting changed.

Outputs are scored by AI judges against a rubric written once, without the judges knowing which variant produced them; a panel of three judges from different vendors can be averaged while per-judge scores stay visible. Experiments end in a scoreboard showing score, flawless outputs, cost per output and latency per variant, which can be exported as a PDF report. Experiments can be re-run and compared side by side, a router can be trained from a finished scoreboard to pick a model per input, and any flow can be deployed as an authenticated REST API with documentation, code examples, batch processing and monitoring.

Key features

• Drag-and-drop flow editor for prompts, scraping, extraction, tools and agents
• Runs one flow across OpenAI, Google, Anthropic and open-source models
• Blind AI judges score outputs against a once-written rubric
• Scoreboard shows score, cost per output and latency per variant
• Trains a per-input model router from experiment results
• Deploys any flow as an authenticated REST API with batch processing

Evaligo at a glance