Promptise Foundry

Promptise Foundry is an open-source framework from Promptise (Manser Ventures) for writing AI agents. Install the promptise package with pip, then use the command line or the Python API to combine language models, MCP tools, memory, approvals, caching and tracing. It connects to OpenAI, Anthropic, Azure, Gemini or Vertex, Bedrock, Mistral and Groq, or to Ollama on your own machine. What the framework calls Skills, such as web_researcher and code_reviewer, are ready-made Python nodes you can combine. What sets it apart is that human approval, multi-tenant isolation and personal-data filtering, the things you worry about once an agent goes live, are part of the framework's design. These are the project's own design descriptions, and we haven't done a security audit.

Who it's for:Developers building AI agents in Python who plan to put them into real business use. If you don't write code, you won't need it.

Before you start: You need Python 3.10 or later, with package metadata listing up to 3.13, and the docs are in English. The official site says the open-source framework needs no account. Prompts, tool results and retrieved content go to whichever model you configure. Where memory, Q&A records and trace logs live depends on your deployment and isn't necessarily local. You pay for model calls and servers yourself. The code is under the Apache-2.0 licence, and each model and outside service has its own terms. Tool permissions, prompt injection and log leaks are yours to control when you deploy.

Based on the project's public documentation. Not tested by AI Inspiration World.