Skip to main content
ZICQ

Topics Agent Frameworks

Agent Frameworks

AI Agent frameworks compared: LangGraph, AutoGen, CrewAI, Letta — design philosophy, tool use model, and production fit.

· Aug 15, 2026 #Agent#agent#LangGraph#LangChain#AutoGen#CrewAI#Letta · 30 / 4345 articles

News · 2026-11-13

AWS Routes Bedrock Coding Agents by Stage with NeMo Switchyard

AWS has published a reproducible reference architecture for deploying LiteLLM with NVIDIA NeMo Switchyard on Amazon ECS Fargate and connecting it to models hosted by Amazon Bedrock. The Stage strategy examines recent tool activity, sends uncertain requests to a lower-cost model, and escalates to a more capable model when failure signals appear. A weighted Random strategy supports traffic splitting and A/B tests. The sample also measures cost per solved task rather than cost per token, reflectin

News · 2026-11-13

NVIDIA Open-Sources SimReady Assets for Robot Pick-and-Place Simulation

NVIDIA has published SimReady assets and a hands-on workflow for building a robot pick-and-place simulation. The tutorial covers scene construction, robot and object setup, goal-pose generation, collision handling, and task execution, providing a reusable path from simulation setup to policy training. The main contribution is engineering infrastructure rather than a new learning algorithm. These assets can reduce setup friction for synthetic-data generation, Sim2Real workflows, and reproducible

News · 2026-11-13

ChatGPT Wallet Lets AI Agents Shop and Check Out for Users

OpenAI has introduced ChatGPT Wallet, allowing AI agents to shop and complete purchases on a user's behalf. The initial rollout is limited to ChatGPT Pro subscribers in the United States. The feature represents more than a conventional payment integration: it extends agentic workflows from product discovery into transaction execution, while raising the stakes around credential isolation, authorization boundaries, and confirmation for irreversible actions. OpenAI has not yet disclosed supported

News · 2026-11-13

PRaVaSS Preserves Uncertain State Beliefs for More Robust Vision-Language Agent Planning

Vision-language agents commonly collapse uncertain perception outputs into one best-guess state, making downstream planning brittle when that guess is wrong. This paper introduces a three-layer factored belief system and a θ-confidence Minimal Belief Set, or MBSθ, that retains only the smallest set of symbolic states whose cumulative probability reaches a user-defined belief-mass threshold. The proposed Probability-Ranked Valid State Search algorithm, PRaVaSS, exploits the structure of this beli

News · 2026-11-13

Simulate What-If Users Before Shipping With Persona-Based Agent Experiments

This work presents a framework for persona-based user simulation that formalizes target cohorts, experimental stimuli, and decision-relevant metrics. Its execution infrastructure runs concurrent experiments across surveys, websites, chatbots, and OS applications, with auditable trial logs, cohort-level aggregation, and cross-variant statistical reporting. The authors evaluate the framework on 12 industry cases spanning five domains. Cross-model comparisons largely preserve primary conclusions,

News · 2026-11-13

Time Is the Signal: Humans Outperform AI at Adaptive Social Coordination

Researchers evaluated 150 human participants, computer-use agents, and text-prompted language models in The Mind, a cooperative card game in which explicit communication is prohibited and timing is the only available signal. Human behavior was best explained by a Bayesian strategy that inferred a partner's beliefs from pause duration and card-selection timing. Frontier models produced broadly compatible inferences, but none efficiently adapted its behavior to the partner over repeated interactio

News · 2026-11-13

SynthCorp Generates Enterprise Data-Agent Benchmarks Through Traceable Event Simulation

SynthCorp creates customizable data estates and benchmarks for LLM-driven data agents by simulating how business events generate and modify enterprise data over time. Each estate combines relational databases, a document store, and a data catalog, accompanied by analytical tasks and executable gold programs. The platform compiles scenarios into event programs, schedules routine and disruption events on a virtual calendar, materializes the resulting state, and records every operation in a causal

News · 2026-11-13

Cleaning Agent History Can Backfire When the World Still Contains the Error

This study separates an agent's conversational history from the state of the external world and shows that sanitizing history does not necessarily improve recovery. On AgentCE-Bench, adding a truthful note beside a preserved error raised the share of correctly completed remaining targets from 51% to 66%, whereas merely recording a completed action reduced it from 50% to 44%. On AppWorld, the effect of a success receipt depended on whether the corresponding database effect actually existed: witho

News · 2026-11-12

Unfinished-Bench Reveals How Persistent Agents Amplify Rule Violations

Unfinished-Bench measures a failure mode that single-turn instruction-following evaluations tend to miss: a persistent agent may continue pursuing an incompatible objective after reading an organizational rule, without a user available to resolve the conflict. Across 97 scenarios spanning seven organizational settings and seven risk categories, the benchmark contrasts persistent and non-persistent execution, as well as user-stated rules with rules embedded in internal documents. Persistent exec

News · 2026-11-12

Expert Labels Can Trigger Overcorrection in LLM Decision Revision

A controlled two-pass audit fixes the question, evidence, and initial model answer, then changes only the revision context: neutral reflection, anonymous disagreement, or disagreement attributed to a domain expert. Across five models and 1,989 evidence-grounded decisions in medicine, science, and law, the study derives an exact selectivity threshold for binary decisions: revision improves accuracy only when the conditional rate of correcting initially wrong answers exceeds the conditional rate o

News · 2026-11-12

Coordinating LLM Inference and Environments Cuts Agent Batch Makespan by Up to 65%

The paper treats an LLM agent batch as a pipeline that alternates between model inference and environment interaction. Its coordinated design combines workload-aware memory planning for hybrid models, trajectory-aware routing for cache reuse, and decoupled sample admission with Chain-Advancing Action Scheduling, or CAAS, to advance long action chains and reduce the batch tail. Controlled replay is used to isolate individual optimizations. Live evaluations with Ling-3.0-Flash and Ling-3.1-Flash r

News · 2026-11-12

LangChain Explains Eval-Driven Development for Reliable AI Agents

LangChain introduces Eval-Driven Development (EDD) for AI agents: define desirable behavior as measurable criteria, write evaluations before implementation, and use repeatable experiments to compare changes. The approach replaces subjective prompt-and-run debugging with a regression-oriented engineering loop. Teams should combine task success, tool-call correctness, latency, and cost, while reserving holdout datasets to detect overfitting. Fixed metrics still cannot capture every failure in lon

News · 2026-11-12

DeepStress Dynamically Corrupts Evidence to Test Deep Search Agent Reliability

DeepStress is a dynamic stress-testing framework for deep search agents. It replaces the retrieval stack with a synthetic environment that generates documents for each query while independently controlling their trustworthiness, relevance, and factuality. This makes evidence degradation reproducible instead of relying on rare failures buried in static benchmark corpora. The framework introduces a five-way outcome taxonomy and two reliability-aware metrics, RAS and CoPRAS, that reward cautious b

News · 2026-11-12

Natera Builds SWARM Agentic AI for Clinical Lab Operations on Bedrock AgentCore

Natera and AWS have deployed SWARM, an agentic AI system for clinical laboratory operations, on Amazon Bedrock AgentCore. Its architecture separates two execution modes: swarm-agent performs autonomous, high-throughput diagnosis for incoming Jira tickets, while swarm-hive supports engineers with interactive troubleshooting, image uploads, and persistent conversation context. The system combines a versioned library of 33 deterministic and LLM-powered skills, an equipment-aware RAG pipeline, even

News · 2026-11-12

Amplitude's Wave Turns Product Signals into a Closed-Loop Coding Agent

Amplitude has built Wave, a proactive product agent primarily powered by OpenAI models. The system identifies what should be improved, coordinates with coding agents to implement changes, and measures the resulting product impact. Its central technical value is not code generation alone, but the creation of a closed workflow spanning product analytics, agentic implementation, release, and measurement. The reported one-to-two-day shipping cycle suggests a tighter product iteration loop, but the

Info · 2026-11-12

GREP.AI Open-Sources AgentRun for Managed, Observable, Self-Improving Agents

GREP.AI has open-sourced AgentRun, the internal agent harness it used to evolve into a managed agent platform. AgentRun connects to existing harnesses—including Pi, Claude Agents, Codex, and OpenCode—through ACP, allowing teams to bring their own agents while adding web-grounded workflows, isolated sandboxes and computer environments, tool integrations, and behavioral analytics. It can be self-hosted with a one-click Render deployment and an OpenRouter key, or used through the hosted console. T

News · 2026-11-12

AWS Updates AgentCore Runtime with 15x Faster Startup and Data-Bound Isolation

AWS has updated the managed Amazon Bedrock AgentCore runtime for long-running, highly autonomous AI agents. Each session still executes inside a hardware-isolated Firecracker MicroVM, but the environment now supports up to 8 vCPUs, 32 GB of memory, and container images of up to 10 GB. AWS reports startup times below two seconds, independent of image size and concurrency. The runtime also reclaims memory shortly after agents release it and can scale idle sessions to zero. It can dynamically moun

News · 2026-11-12

AWS Adds a Coding-Agent Skill for Migrating Bedrock Managed Knowledge Bases

AWS has added a knowledge-base migration guide to the `amazon-bedrock` skill in Agent Toolkit for AWS. Coding agents can inspect an existing customer-managed Bedrock knowledge base through read-only calls, map its configuration, and produce either a live provisioning plan or a migrated CloudFormation, CDK, or Terraform template. Resource creation remains behind explicit approval, while ingestion, retrieval validation, and traffic cutover stay with the application team. The guide also documents

News · 2026-11-12

AWS Connects Cross-Organization A2A Agents with Bidirectional OAuth

AWS has published a reference architecture for cross-organization agent collaboration using the Agent-to-Agent (A2A) protocol. A LangChain orchestrator running on Amazon ECS discovers and invokes a Strands agent deployed on Amazon Bedrock AgentCore through an Agent Card, while both sides authenticate through separate Cognito pools and support bidirectional OAuth-protected callbacks. The design isolates internal tools, prompts, models, and SDKs behind a protocol-level interface based on JSON-RPC

News · 2026-11-12

AWS Builds Production-Ready Autonomous Agents with Spring AI and Bedrock AgentCore

AWS has published a production-oriented reference implementation for building autonomous agents in Java with Spring AI and Amazon Bedrock AgentCore. The samples cover tool calling, conversational and long-term memory, resilience, deterministic rules, multi-agent routing, and deployment through Terraform and an ARM64 container. The architecture separates application-level Spring components from model- and framework-agnostic AgentCore services for runtime, memory, browser automation, code executi

News · 2026-11-11

Coding Agents Solve Early but Keep Working: Termination Lag Consumes 35–60% of Steps

The paper Termination Lag measures how long coding agents continue working after they have already produced a correct repository patch. Across five models, two agent frameworks, and three SWE-bench datasets, the authors recorded and tested every distinct patch generated during each run. They found that 38–82% of successful runs solved the task with their initial source-code edit, yet Qwen3-Coder devoted a median 35–37% of its successful-run steps to post-solution activity; for Laguna-XS.2, the f

News · 2026-11-11

Continual-ARC Evaluates Lifelong Learning: Harness Memory Trails Weight Updates

Continual-ARC reorganizes three ARC-AGI editions into a lifelong task stream. Previously encountered tasks return with fresh instances after dormant gaps while new tasks continue to arrive. Both memory-based and weight-based learners consume a shared feedback budget covering demonstrations, harness actions, submissions, and paid retries. After the stream ends, the learner is frozen and evaluated separately on fresh instances of encountered tasks for memorization and on unseen tasks for generaliz

News · 2026-11-11

JustRUA Isolates What Actually Makes Robot-Use Agents Work

JustRUA evaluates robot-use agents within a unified harness that isolates the effects of full episode history, action interfaces, experience reuse, scene geometry, and early motion termination. Its analysis finds that retaining complete trajectory history and stopping motions early account for much of the improvement, while experience reuse introduces trade-offs rather than free gains. By adding geometric scene information—even depth predicted from RGB—the lightweight harness raises the RoboDojo

Info · 2026-11-11

Klad Lets Coding Agents Paint With Editable, Layer-Aware Brush Strokes

Klad is a Rust-built painting application designed for collaboration between coding agents and human artists. Its CLI exposes brushes, frame-by-frame animation, vectors, text layers, colors, layer effects, and unlimited undo, allowing an agent to create artwork that remains structurally editable as strokes and layers rather than a flattened bitmap. The file format, CLI reference, and agent skill are released under the MIT License, while the main application is source-available. Klad targets a g

Info · 2026-11-11

Chrome Launches Modern Web Guidance 1.0 for AI Coding Agents

Google Chrome has released Modern Web Guidance 1.0, a curated knowledge resource that gives coding agents better grounding in modern web-platform capabilities, browser compatibility constraints, progressive enhancement, accessibility, and performance practices. Its main value is not a new model capability. Instead, it turns frequently changing browser standards into contextual guidance that agents can retrieve while generating and reviewing code, reducing the likelihood of obsolete APIs, broken

Info · 2026-11-11

AWS Open-Sources an Emulator for Autonomous Cloud-Application Development

AWS has open-sourced aws-emulator, a programmable environment that emulates AWS services so coding agents can generate infrastructure code, exercise deployment workflows, and validate the resulting state without operating directly on a production account. The project turns a conventional infrastructure-as-code workflow into an iterative generate–deploy–verify loop. Its practical value lies in reducing cloud spend, limiting accidental changes, and giving agents a faster feedback surface. Product

News · 2026-11-11

AWS Automates Agent Registry Metadata Sync with Event-Driven Architecture

AWS has published a deployable push-based pattern for keeping AWS Agent Registry synchronized with MCP server changes. CloudTrail captures Amazon Bedrock AgentCore runtime updates, Amazon EventBridge routes those events across accounts, and AWS Lambda retrieves the current tool definitions, compares their schemas, and updates matching registry records. AgentCore Identity supplies scoped OAuth tokens without storing client secrets in Lambda. The design preserves governance by returning modified

News · 2026-11-11

DXC OASIS Deploys Governed Multi-Model AI Agents on Amazon Bedrock

DXC Technology has built the multi-model substrate for DXC OASIS on Amazon Bedrock, separating governed model access from the orchestration and assurance harness that controls agent actions. The architecture uses Amazon EKS in private subnets, AWS PrivateLink, IAM, KMS, Bedrock Guardrails, tiered permissions, approval gates, and OpenTelemetry traces to support regulated production workloads. DXC reports that OASIS reduced average resolution time by 95% across 56 measured use cases. The case ill

News · 2026-11-11

LlamaParse Pen Tests Against OWASP LLM Top 10, Uncovering Hidden-Text Extraction Risks

LlamaIndex commissioned Trace to run a four-day agentic dynamic application security assessment of LlamaParse. The grey-box test covered parsing, classification, field extraction, indexing, retrieval, deployed agents, and the MCP server against all ten OWASP Top 10 risks for LLM applications. One notable finding involved a deliberately crafted invoice with white-on-white hidden lines. Although a human reader saw only the ordinary invoice, the extractor returned altered totals and banking detail

Info · 2026-11-11

Klad Open-Sources an Agent-Controlled Paint App with Editable Strokes

Klad is a Rust-based paint application designed for coding agents to operate through a command-line interface while preserving human editability. Its central design choice is to represent editable brush strokes, layers, animation, vectors, text, and effects as structured document state rather than reducing the result to an immutable generated bitmap. This allows an agent to produce a first draft and users to revise individual strokes through the graphical interface. The project also provides in