Tool Use

OpenClaw v2026.7.1-beta.2 introduces GPT-5.6 support, external harness attachment via `openclaw attach`, Telegram Codex workflows with login and steering, event-driven cron runs with `on-exit` scheduling, and native app improvements including iOS 26 visual system adoption. The release also adds ClawRouter for dynamic model discovery with credential-scoped routing, richer iMessage messaging with native polls, safer scoped conversations via capability profiles, and enhanced Control UI navigation with reasoning-effort controls. Extensive fixes address Telegram durability, agent reliability, provider safety, channel delivery, cron correctness, and memory recovery across multiple platforms.

Claude Code v2.1.201 is a minor release that improves Claude Sonnet 5 session handling by removing mid-conversation system role usage for harness reminders. This change streamlines the conversation flow and reduces unnecessary system interventions during active coding sessions.

Claude Code v2.1.200 introduces significant stability and usability improvements, including changes to user interaction defaults (manual permission mode, no auto-continue dialogs), fixes for background session crashes and daemon management issues, and enhancements to accessibility features. The release addresses critical bugs in MCP server handling, subagent rate limiting, and terminal rendering, while improving screen reader support and installation robustness.

Claude Code v2.1.199 is a maintenance release focused on reliability and user experience improvements. Key fixes address SSL certificate handling, streaming response preservation, subagent error reporting, background agent stability on Linux and macOS, and UI/UX issues like progress indicators and agent panel display. The release also improves retry logic for transient server errors and fixes several edge cases in session management and tool routing.

OpenClaw v2026.7.1-beta.1 introduces GPT-5.6 model support, external harness attachment via `openclaw attach`, Telegram Codex workflow integration, event-driven cron scheduling with `on-exit` triggers, and native app modernization across iOS and Android. The release includes enhanced messaging capabilities (iMessage polls, Telegram pairing), safer scoped conversations with capability profiles, and extensive reliability fixes across Telegram durability, agent context, provider safety, and memory recovery. Over 100 contributors improved diagnostics, channel routing, Windows execution, and mobile UI stability.

Claude Code v2.1.198 introduces Claude in Chrome as generally available, adds background agent notifications and a new `/dataviz` skill for chart design, and expands upstream provider support with Claude Platform on AWS. The release includes significant improvements to agent workflows, including automatic PR creation for code work, better error handling with transient retry logic, and inheritance of session configurations by subagents. Multiple bug fixes address agent lifecycle management, network resilience, UI rendering, and authentication token refresh.

This cookbook demonstrates how to reproduce Claude's published agentic search benchmark scores (DeepSearchQA, BrowseComp) using the Messages API with programmatic tool calling, server-side compaction, and task budgets. The key is proper harness configuration—API parameters that become critical for agents running 30+ tool calls across hundreds of thousands of tokens. By following this guide, you'll build an agentic search loop that matches Claude's official benchmark performance and understand why each configuration choice matters for long-horizon tasks.

crewAI 1.15.2a1 release introduces inline skill definitions, stream frame protocol for flows, and enhanced type support in CrewDefinition. Key improvements include repointing template commands to crewAIInc-fde org and adding a generated Flow Definition authoring skill. Bug fixes address documentation navigation issues, with updates to cost limit rules and logging configuration guides.

Claude Code v2.1.197 introduces Claude Sonnet 5 as the new default model, featuring a native 1M-token context window for enhanced processing capabilities. The update includes promotional pricing of $2/$10 per million tokens through August 31, making it more cost-effective for users. This release represents a significant upgrade in model performance and context handling for the Claude Code platform.

OpenClaw v2026.6.11 is a major release featuring enhanced channel control across Slack, Mattermost, and WhatsApp with relay modes and per-DM model overrides. The update introduces richer operator workflows through file-driven message handling and RAFT CLI wake bridges, safer plugin distribution with externalized official plugins, stronger mobile operations via Android settings improvements, and more reliable agent turns through Codex partial deltas and prompt-cache stability. The release includes 308 merged PRs addressing channel delivery reliability, gateway safety, provider model coverage, and configuration guardrails across multiple messaging platforms.

Claude Code v2.1.196 introduces organization-level default model settings, improved session management with readable names, and enhanced security for MCP server handling. The release includes numerous bug fixes for background jobs, PowerShell compatibility, agent panel functionality, and streaming reliability. Key improvements focus on session persistence across restarts, better file attachment handling, and reduced token usage in code review workflows.

langchain-openrouter version 0.2.5 has been released with three key fixes. The release addresses deduplication of repeated finish metadata, strips Response reasoning IDs, and includes general release improvements. This is a patch update from version 0.2.4 that focuses on bug fixes and data cleanup for OpenRouter integration.

OpenClaw v2026.6.11-beta.2 introduces enhanced channel control with Slack relay mode and native Mattermost commands, richer operator workflows via file-driven message handling and remote wake-up paths, safer plugin distribution with externalized plugins and icon metadata, stronger mobile operations with improved Android settings, and more reliable agent turns through Codex partial deltas and prompt-cache stability. The release includes 308 merged PRs addressing gateway tooling, provider coverage, channel delivery reliability, WhatsApp message identity preservation, gateway session safety, agent fallback behavior, and configuration guardrails across multiple integrations.

CrewAI 1.15.1 release introduces Git repository initialization for generated projects, requires explicit project definitions, and enhances deployment workflows. The update includes critical bug fixes for JSON rendering, SSRF vulnerabilities, and deployment page linking. Documentation improvements focus on open source positioning and agent setup guidance.

langchain-anthropic version 1.4.8 was released with bug fixes and dependency updates. Key improvements include preserving initial text during content block start events, upgrading langgraph-checkpoint to 4.1.1, and adding error messages to bare ValueError calls in core. This patch release enhances stability and error handling in the Anthropic integration for LangChain.

langchain-fireworks version 1.4.3 is a maintenance release focused on dependency updates and stability improvements. The release includes bumps to vcrpy, langsmith, pytest, and other core dependencies, along with model profile data refreshes and documentation updates. Key improvements include package version tracking for tracing metadata, type-checking enhancements, and compatibility fixes for aiohttp and vcrpy.

Claude Code v2.1.193 introduces enhanced auto-mode shell command classification, improved OpenTelemetry logging with assistant response tracking, live file path autocomplete in bash mode, and MCP server authentication notifications. The release fixes critical issues with background agent management, session persistence, and UI state handling while improving background task memory management and plugin auto-renaming functionality.

crewAI 1.15.0 introduces significant enhancements to conversational flows and declarative configuration, including unified flow loading, DMN mode support, and CLI TUI integration. The release adds new composite actions (each, single agent, crew actions) to FlowDefinition and improves telemetry tracking for conversational turn usage. Multiple bug fixes address JSON schema handling, credential permissions, token aggregation, and symlink vulnerabilities, while documentation improvements include Datadog integration guides and JSON-first crew project setup.

OpenClaw v2026.6.11-beta.1 introduces enhanced channel control with Slack relay mode and Mattermost native commands, richer operator workflows via file-driven message handling and remote wake-up paths, safer plugin distribution with externalized plugins and icon metadata, stronger mobile operations with improved Android settings, and more reliable agent turns through Codex partial deltas and prompt-cache stability. The release includes 305 merged PRs addressing channel delivery reliability, WhatsApp message identity preservation, gateway session safety, agent fallback behavior, provider model edge cases, and configuration guardrails across Telegram, WhatsApp, Slack, Mattermost, and other integrations.

OpenAI Agents Python SDK v0.17.7 introduces configurable WebSocket limits, buffered Chat Completions tool-call streaming, and numerous stability improvements. Key fixes address sandbox buffering, message cleanup, database atomicity, guardrail task cancellation, and multi-agent tool dispatch. The release also improves error handling, PTY output collection, and documentation with contributions from 10 new community members.