EarthLink Launches Prompt Flow Studio, With Enterprise Controls Planned for October

EarthLink Network has launched Prompt Flow Studio worldwide on the web, putting model comparison, AI-to-AI debate and collaborative drafting in one workspace. The core product is available now; its planned sensitive-data controls and audit features are due to begin a phased rollout in October 2026.

The product is designed for work that moves among AI vendors rather than staying with one model. EarthLink says it supports more than 50 providers and more than 400 models across text, image, audio and video, including access through OpenRouter and fal.ai.

Comparison is the starting point

Users can send one prompt to several models and view outputs side by side, including images generated simultaneously by multiple image models. Per-model costs are displayed with those results. Personas can define a model’s expertise, stance and tone, including opposing roles assigned to the same underlying model.

Debate turns that setup into a structured exchange: models and personas can serve as advocate, critic, facilitator and red-team reviewer before another model synthesizes an answer. Co-Draft instead has several AIs propose outlines, review one another’s work and send it to an integrator model to produce versions of one document.

50+AI providers

EarthLink says Prompt Flow Studio supports more than 50 AI providers.

400+Models

The company says the workspace includes more than 400 models across text, image, audio and video.

One document base, different model choices

The workspace also offers model-agnostic retrieval-augmented generation, or RAG, which lets users upload documents and use them as source material with a chosen model. EarthLink says the same document context can be used with models from OpenAI, Anthropic, Google and others, either in a documents-only mode or a blended mode. A handover feature carries objectives, constraints and conclusions when a user changes models.

The governance layer remains scheduled work

Starting in October, EarthLink plans to roll out enterprise controls that block sensitive information before it reaches an external AI service or automatically mask it and restore the original values in the returned response. The roadmap also includes custom detection rules and controls over permitted AI services and models.

The planned features also include hash-chained audit logs, legal hold, retention of records for up to seven years and selective deletion of personal data. EarthLink says administrators would not have routine access to prompts or AI responses, while legal disclosures would follow a controlled, audited process. The timing and phased delivery of that layer are the material constraint for organizations evaluating the product for sensitive work.

The free tier includes the main workspace functions with lightweight models, 1 GB of RAG storage and 300 PFC credits at signup; paid plans unlock all models. iOS and Android apps are rolling out after the web launch.