Context Management

OpenClaw v2026.4.14-beta.1 is a maintenance release addressing 30+ security fixes, stability improvements, and feature enhancements across messaging platforms, browser automation, scheduling, memory systems, and agent tooling. Key improvements include replacing markdown parsing to prevent ReDoS attacks, fixing Telegram forum topic handling, correcting cron scheduler behavior, and enforcing SSRF/security policies across multiple integrations. The release emphasizes background task optimization and proper context preservation across system components.

The Claude Agent SDK provides built-in session management functions to build conversation history sidebars without writing custom parsers. This cookbook demonstrates how to list, read, rename, tag, and fork sessions stored as JSONL transcripts on disk. By leveraging SDK primitives like list_sessions(), get_session_messages(), and rename_session(), developers can implement session browsers for desktop apps, IDE extensions, and chatbots without managing file I/O directly.

This cookbook teaches context engineering strategies for long-running AI agents, focusing on three key techniques: compaction (summarizing context), tool-result clearing (removing re-fetchable tool outputs), and memory (persistent external storage). The guide addresses context rot—the degradation of model performance as context windows grow—and provides practical implementations using Claude's API. Through a research agent example, it demonstrates how to combine these strategies to manage token growth, maintain conversation continuity, and persist knowledge across sessions.

This guide teaches how to build knowledge graphs from unstructured text using Claude's structured outputs for entity extraction, relation mining, and entity resolution. Rather than training traditional NER and relation classifiers, Claude handles each stage via prompts, enabling multi-hop graph reasoning without a database. The approach uses Haiku for high-volume extraction and Sonnet for entity resolution, with techniques transferable to production databases like Neo4j or PostgreSQL.

This tutorial introduces Projects in Claude.ai, a feature that enables users to organize their work by maintaining persistent context and custom instructions across conversations. Projects allow you to group related conversations, documents, and settings in one place, making it easier to manage complex workflows and maintain consistency across multiple interactions with Claude.

Speculative prompt caching is a technique that reduces time-to-first-token (TTFT) by warming up Claude's cache while users are still formulating their queries. Instead of waiting until a user submits their question to load context into the cache, the system begins cache warming immediately when the user starts typing. This cookbook demonstrates the pattern using SQLite source code as example context, showing how to implement it with the Anthropic Python SDK using async operations and prompt caching controls.

This cookbook demonstrates how to build AI agents with persistent memory using Claude's memory tool and context editing capabilities. It addresses challenges of long-running agents losing learned patterns between sessions and context window overflow by implementing cross-conversation learning and automatic context management. The guide covers practical implementations for use cases like code review assistants, research assistants, and customer support bots, with setup instructions and best practices for memory security and organization.

This cookbook teaches developers how to manage long-running Claude conversations by implementing session memory compaction using background threading and prompt caching. Rather than waiting for context limits to be exceeded (reactive approach), the pattern enables instant compaction by proactively building summaries in the background. The guide covers writing effective session memory prompts, implementing background threading for zero-latency compaction, and applying prompt caching to reduce costs by ~80%. It includes Python code examples demonstrating both traditional (slow) and instant (fast) compaction strategies for conversational applications.

This cookbook demonstrates automatic context compaction for managing token limits in long-running agentic workflows. It shows how the Claude Agent Python SDK can automatically compress conversation history when token usage exceeds a threshold, enabling tasks to continue beyond the 200k token context limit. The example uses a customer service agent processing support tickets, where each ticket requires multiple tool calls that accumulate in conversation history. By implementing context compaction with the compaction_control parameter, agents can maintain focus and efficiency across many iterations without manual context management.

This cookbook demonstrates how to scale Claude applications from dozens to thousands of tools using semantic embeddings for dynamic tool discovery. Instead of front-loading all tool definitions (which consumes context and increases latency), the approach provides Claude with a single tool_search tool that returns relevant capabilities on demand, reducing context usage by 90%+. The guide walks through implementing client-side semantic search using SentenceTransformer embeddings to match user queries with appropriate tools from large libraries, making it practical for production applications managing extensive tool ecosystems.

OpenClaw v2026.3.7 introduces major enhancements to context engine plugins, persistent channel bindings for Discord/Telegram, multi-language UI support (Spanish), improved web search capabilities, and Docker containerization optimizations. The release includes a breaking change requiring explicit gateway auth mode configuration. Key additions enable alternative context management strategies, durable ACP thread routing, per-topic agent overrides, and enhanced security with config validation.

OpenAI Agents Python SDK v0.9.0 drops Python 3.9 support and introduces configurable timeout handling for function tools with options for timeout_seconds, timeout_behavior, and timeout_error_function. The release also narrows the Agent#as_tool() return type to FunctionTool, adds ToolOutputTrimmer for context management, and includes fixes for tracing and audio buffer handling.

OpenClaw v2026.2.23 release adds first-class Kilo Gateway and Moonshot provider support, introduces prompt caching documentation and per-agent parameter overrides, and hardens session maintenance with disk-budget controls and security headers. The release includes numerous fixes for browser SSRF policies, Telegram polling/reactions, context overflow detection, and agent reasoning/compaction behavior.