OpenAI Agents SDK
PulseAugur coverage of OpenAI Agents SDK — every cluster mentioning OpenAI Agents SDK across labs, papers, and developer communities, ranked by signal.
- 2026-08-11 product_launch OpenAI Agents SDK version 0.20.0 was released, updating the default model and introducing new features. source
6 day(s) with sentiment data
-
Pexafy launches tool for AI agents to find real stock photos
Pexafy has launched a new tool that allows AI agents to search for and retrieve real photographs from various stock photo libraries. The tool integrates with popular AI frameworks like LangChain and Vercel AI SDK, enabl…
-
AI agents can refuse actions verbally but execute them via tools, new tests reveal
A blog post highlights a critical flaw in AI agent evaluation where agents can verbally refuse an action while simultaneously executing it through tool calls. The author proposes a testing methodology that logs all tool…
-
AI agents lack per-call budget checks, creating cost risks
The article discusses the limitations of current budget checks for AI agents, particularly within OpenAI's Agents SDK and similar frameworks. It highlights that while OpenAI has monthly spending limits, these are not en…
-
LLM agent frameworks show audit log gaps in parent-child handoffs
A technical analysis of five LLM agent frameworks reveals inconsistencies in how audit logs are maintained during parent-child agent handoffs. While the hash-chained audit log generally preserves the child agent's tool …
-
Agent harnesses, frameworks, and MCP: Understanding AI agent architecture layers
The article distinguishes between three distinct layers in AI agent architecture: agent harnesses, agent frameworks, and the Model Context Protocol (MCP). Agent harnesses, like those in OpenAI's Codex and Anthropic's Cl…
-
AI agent frameworks lack critical audit fields for evidence, study finds
A recent audit of six popular AI agent frameworks revealed significant gaps in their logging capabilities for evidentiary purposes. On September 4, 2026, researchers found that the median framework records only five out…
-
FreshCtx 0.11.0 adds verifiable action trails for AI agents
FreshCtx 0.11.0 has been released, introducing a method for AI agents to create verifiable records of their consequential actions. This new version focuses on establishing an evidence-to-action trail, allowing developer…
-
LangGraph, OpenAI, and Claude Agent SDKs Compared for AI Agent Development
This article compares three leading agent frameworks: LangGraph, OpenAI Agents SDK, and Claude Agent SDK. It analyzes their distinct approaches to managing agent loops and solving similar problems with different core pr…
-
AWS Bedrock AgentCore Evaluations uses OpenTelemetry for framework-agnostic AI agent assessment
Amazon Bedrock AgentCore Evaluations offers a new way to assess AI agent frameworks by decoupling evaluation from specific framework choices. It leverages OpenTelemetry, a vendor-neutral instrumentation standard, to col…
-
OpenAI Agents SDK 0.20.0 updates default model to GPT 5.6 Luna
The OpenAI Agents SDK has been updated to version 0.20.0, which implicitly changes the default model to GPT 5.6 Luna. This update affects applications that relied on the previous implicit default model, altering their r…
-
AI agent summarization silently loses rules, increasing policy violations
A new benchmark called ConstraintRot, detailed in the paper Governance Decay, reveals that AI agent summarization techniques can silently lose critical rules, leading to increased policy violations. When rules were full…
-
Control agent turns, not models, to cut LLM costs
A common issue in multi-agent AI systems, leading to excessive API costs and broken workflows, is the tendency for agents to enter conversational loops or excessively hand off tasks. The author argues that instead of sw…
-
Scrapeless server grants OpenAI Agents SDK live web access
Scrapeless has introduced a new server that grants OpenAI Agents SDK agents live web access. This integration allows agents to utilize a suite of 21 tools, including web scraping, Google searches, and browser control fu…
-
LLM agents gain browser control with Playwright MCP
Playwright MCP offers a practical solution for integrating LLM agents with live web browsers, enabling them to perform actions like navigating, clicking, and filling forms. This tool acts as the agent's interface to a r…
-
Developer details secure Python sandbox for LLM agents using Docker
A developer has detailed a method for enabling LLM agents to execute Python code within a secure Docker sandbox. This approach moves beyond simple code generation by creating an execution loop where the agent writes cod…
-
AWS Bedrock AgentCore adds web search for grounded AI agents · 1 source tracked
AWS has made Web Search on Amazon Bedrock AgentCore generally available, allowing AI agents to access current, cited web results without data leaving the AWS environment. This managed tool, priced at $7 per 1,000 querie…
-
CrewAI vs. AutoGen: Choosing the Right Multi-Agent Orchestration Framework
The article compares two leading frameworks for orchestrating multi-agent AI systems: CrewAI and AutoGen. CrewAI is presented as ideal for structured, predefined workflows where roles and tasks are clearly defined, akin…
-
Top 5 frameworks lead agentic AI development for autonomous systems
Several key frameworks are emerging as leaders in the development of agentic AI systems, with a focus on enabling autonomous, multi-step processes. LangGraph, CrewAI, OpenAI Agents SDK, Google ADK, and Mastra are highli…
-
Knowledge Graphs and Vector DBs Combine to Enhance AI Agent Memory
Researchers are exploring new methods to improve knowledge graph completion (KGC) by addressing the limitations of traditional triplet prediction. One approach introduces a relation set completion task (RSC) to infer se…
-
AI code agents use rule files for consistent project-wide code generation
AI code generation tools like Claude Code and Cursor can be guided by project-specific rule files, such as CLAUDE.md and AGENTS.md, to ensure consistent code style and adherence to best practices. These files, placed at…