psi.run positions itself as an AI Agent design and runtime workbench. Its site copy emphasizes βyou design the Agent, then step back and let the universe unfold,β meaning users first define an Agentβs boundaries, rules, and skills, then observe it in action. The currently captured content shows entry points such as Workbench, Owner profile, Agent IP Square, Live Square, and Main Stage, with the core functionality centered on custom Agent creation.
In terms of AI capabilities and models, psi.runβs main strength is its broad coverage of model providers. The page lists OpenAI, Anthropic, Google Gemini, DeepSeek, xAI, OpenRouter, Groq, Mistral AI, Together AI, Perplexity, Cohere, Alibaba Qwen, Moonshot Kimi, Azure OpenAI, and others. It also supports Local / Ollama, Private model, and OpenAI-compatible API endpoint. Users can configure a Primary model and Secondary model, and enter core skills with a 500-character limit. Overall, it feels more like a multi-model Agent configuration console than a standalone chat tool.
The captured page does not provide details on free quotas, subscription pricing, pay-as-you-go billing, or enterprise plans, nor does it specify supported payment methods. On privacy, the only visible statement is βAPI key, encrypted in server mode,β indicating that API keys are encrypted when stored in server mode. However, there is no clear information on data retention, logging, whether conversation content is used for training, or permission isolation, so it should still be evaluated carefully before production use.
The main advantage is flexible model integration, especially support for local/Ollama setups, private models, and OpenAI-compatible endpoints, making it suitable for users who want to experiment across multiple models. The Agent design fields are also fairly intuitive, allowing users to set a name, title, IP gender, and core skills. The downside is that public information is limited: there are no concrete output examples, workflow capabilities, plugin ecosystem details, API documentation, service SLA, or pricing information, making it difficult to assess its maturity and long-term operating cost.
psi.run is best suited for developers, AI Agent enthusiasts, and prompt/character designers who want to quickly prototype Agents or compare results across different models. Access from China is not addressed in the captured content, so its availability is unknown; payment methods are also undisclosed. If you need a more mature Chinese-language ecosystem and more controllable deployment options, alternatives such as Dify, Coze, Flowise, Langflow, OpenWebUI, or Ollama are worth comparing.
β This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on psi.run official site.
psi.run is an Unknown AI Apps provider. TG4G tracks its product information, an overall rating of 6.0/10, and a China-accessibility score of Workable. Click "Visit Official Site" to reach psi.run directly.