Simon Willison explores his latest approach to running untrusted Python code safely within applications by utilizing MicroPython inside a WebAssembly (WASM) sandbox. The project addresses the security risks of plugin systems where code normally executes with full privileges, potentially leading to data leaks or system compromise. By leveraging wasmtime and an alpha package called micropython-wasm, Willison demonstrates how to enforce memory and CPU limits while providing controlled access to host functions through a custom thread-based request queue for persistent interpreter state.
Main topics:
- Security challenges in Python plugin systems
- Advantages of WebAssembly as a sandboxing technology
- Building the micropython-wasm alpha package
- Implementation details for persistent state and host functions
- Integration with Datasette Agent to execute code via LLMs
Main topics:
- Security challenges in Python plugin systems
- Advantages of WebAssembly as a sandboxing technology
- Building the micropython-wasm alpha package
- Implementation details for persistent state and host functions
- Integration with Datasette Agent to execute code via LLMs
AI agents operate through a ReAct (Reason + Act) pattern implemented as a deterministic Python `while` loop that maintains conversation history within the context window to serve as short-term memory. The core logic involves sending the system prompt and cumulative tool results to an LLM, which returns either a final answer or structured function calls; if tools are requested, their outputs are executed and appended back into the message list for subsequent reasoning iterations. This architecture supports local execution via Ollama's OpenAI-compatible API, mixed-mode orchestration by delegating complex tasks from local models to cloud APIs through specialized tool functions, and scalable tool integration using the Model Context Protocol (MCP) to dynamically discover and invoke external services via JSON-RPC.
Memori is an agent-native memory infrastructure that acts as an LLM-agnostic layer to transform AI agent execution and conversations into structured, persistent state for production systems. It integrates seamlessly into existing architectures, allowing agents to automatically capture and recall information from past interactions without requiring changes to core code or prompts.
Key features and points:
* Provides advanced augmentation of memories including attributes, facts, preferences, relationships, and skills at the entity, process, and session levels.
* Achieves high accuracy and token efficiency in long-conversation memory as demonstrated by LoCoMo benchmark results.
* Offers dedicated SDKs for both Python and TypeScript.
* Supports Model Context Protocol (MCP) for easy connection to developer tools like Claude Code and Cursor.
* Compatible with a wide range of LLMs including OpenAI, Anthropic, Gemini, DeepSeek, and Grok, as well as frameworks like LangChain and Pydantic AI.
Key features and points:
* Provides advanced augmentation of memories including attributes, facts, preferences, relationships, and skills at the entity, process, and session levels.
* Achieves high accuracy and token efficiency in long-conversation memory as demonstrated by LoCoMo benchmark results.
* Offers dedicated SDKs for both Python and TypeScript.
* Supports Model Context Protocol (MCP) for easy connection to developer tools like Claude Code and Cursor.
* Compatible with a wide range of LLMs including OpenAI, Anthropic, Gemini, DeepSeek, and Grok, as well as frameworks like LangChain and Pydantic AI.
This quickstart guide provides a step-by-step walkthrough for building, testing, and deploying AI agents using the Amazon Bedrock AgentCore CLI.
- code-based agents for full orchestration control using frameworks like LangGraph or OpenAI Agents
- managed harness preview for rapid configuration-based deployment.
- code-based agents for full orchestration control using frameworks like LangGraph or OpenAI Agents
- managed harness preview for rapid configuration-based deployment.
This repository focuses on the concept of an "agent" as a trained model, not just a framework or prompt chain. It emphasizes building a "harness" – the tools, knowledge, and interfaces that allow the model to function effectively in a specific domain. The core idea is that the model *is* the agent, and the engineer’s role is to create the environment it needs to succeed.
The content details a 12-session learning path, reverse-engineering the architecture of Claude Code to understand how to build robust and scalable agent harnesses. It highlights the importance of separating the agent (model) from the harness, and provides resources for extending this knowledge into practical applications.
The content details a 12-session learning path, reverse-engineering the architecture of Claude Code to understand how to build robust and scalable agent harnesses. It highlights the importance of separating the agent (model) from the harness, and provides resources for extending this knowledge into practical applications.
agentic_TRACE is a framework designed to build LLM-powered data analysis agents that prioritize data integrity and auditability. It addresses the risks associated with directly feeding data to LLMs, such as fabrication, inaccurate calculations, and context window limitations. The core principle is to separate the LLM's orchestration role from the actual data processing, which is handled by deterministic tools.
This approach ensures prompts remain concise, minimizes hallucination risks, and provides a complete audit trail of data transformations. The framework is domain-agnostic, allowing users to extend it with custom tools and data sources for specific applications. A working example, focusing on stock market analysis, demonstrates its capabilities.
This approach ensures prompts remain concise, minimizes hallucination risks, and provides a complete audit trail of data transformations. The framework is domain-agnostic, allowing users to extend it with custom tools and data sources for specific applications. A working example, focusing on stock market analysis, demonstrates its capabilities.
Learn how to equip your Microsoft Agent Framework agents with portable, reusable skill packages that provide domain expertise on demand using Agent Skills. This article covers what Agent Skills are, progressive disclosure, creating skills, connecting skills to an agent (with .NET and Python examples), use cases, and security considerations.
ClawWork is an AI coworker framework that transforms AI assistants into agents capable of completing real-world professional tasks and earning income. It features a live economic benchmark, multi-model competition, and rigorous LLM evaluation. The project provides tools for standalone simulation and integration with openclaw/nanobot.
LlamaAgents Builder allows users to build document agents using natural language, generating agent workflows for tasks like classifying financial statements, extracting data from resumes, and creating multi-document summarization pipelines. It offers a balance between low-code ease of use and the flexibility of custom development, generating Workflows that can be deployed on LlamaCloud or self-hosted.
Minimal Claude Code alternative. Single Python file, zero dependencies, ~250 lines.
