pluginadvancedOpenCLAW

OpenClaw v2026.3.7-beta.1: Customizable ContextEngine and Dual-Engine Routing

OpenClaw, a framework for building automated AI assistants, has a major update. It introduces a 'ContextEngine' that lets developers easily customize how the AI remembers and uses information, enabling more powerful and tailored solutions. This release also adds Dual-Engine Routing to automatically switch between AI providers like OpenAI and Google if one fails, significantly improving reliability.

AI SETUP PROMPT

Paste into Claude Code or Codex CLI — it will scan your project and set everything up

# Install & Configure: OpenClaw v2026.3.7-beta.1: Customizable ContextEngine and Dual-Engine Routing

## What This Is
OpenClaw, a framework for building automated AI assistants, has a major update. It introduces a 'ContextEngine' that lets developers easily customize how the AI remembers and uses information, enabling more powerful and tailored solutions. This release also adds Dual-Engine Routing to automatically switch between AI providers like OpenAI and Google if one fails, significantly improving reliability.

Source: https://www.youtube.com/watch?v=CwVP9QUyC3I

## Before You Start

Scan my workspace and analyze:
- The project language, framework, and directory structure
- Existing agent configuration (check for .claude/, .codex/, CLAUDE.md, settings.json, commands/, skills/ directories)

Then ask me before proceeding:
1. Do I need to configure any service credentials for this plugin (database, API keys, etc.)?
2. Should this be project-scoped or global?

## Source Access Note

The source URL (https://www.youtube.com/watch?v=CwVP9QUyC3I) may not be directly accessible from the terminal. Use the Reference Implementation and Additional Context sections below instead. If you need more details, ask me to paste relevant content from the source.

## What to Implement

This is an **Agent Plugin** — typically an MCP server or capability bundle.

- Install any required npm/pip packages from the source repository
- Register the MCP server in `.claude/settings.json` under the `mcpServers` key with the correct command and args
- Configure any required environment variables or credentials
- Verify the server starts correctly by checking its expected output

## Additional Context

- Scan the user's project to identify an existing OpenClaw installation and its version. If an older version is found, plan the upgrade path for the framework's core packages to v2026.3.7-beta.1.
- Inspect the current agent configurations to implement the new Dual-Engine Routing. Create a fallback chain in the agent's settings, using their primary model provider first and configuring one or two secondary providers from their existing API keys.
- Scaffold a new ContextEngine plugin file implementing the core lifecycle hooks (`bootstrap`, `ingest`, `assemble`, `compact`). Based on user's goal, implement a basic RAG strategy by having the `ingest` hook add document embeddings to a vector store and the `assemble` hook retrieve context for the prompt.

## Guidelines

- Adapt everything to my existing project — do not assume a specific stack or directory layout
- Review any fetched code for safety before installing or executing it
- After setup, run a quick verification and show me a summary of exactly what was installed, where, and how to use it

2,757 charactersCompatible with Claude Code & Codex CLI

MANUAL SETUP STEPS

01Scan the user's project to identify an existing OpenClaw installation and its version. If an older version is found, plan the upgrade path for the framework's core packages to v2026.3.7-beta.1.
02Inspect the current agent configurations to implement the new Dual-Engine Routing. Create a fallback chain in the agent's settings, using their primary model provider first and configuring one or two secondary providers from their existing API keys.
03Scaffold a new ContextEngine plugin file implementing the core lifecycle hooks (`bootstrap`, `ingest`, `assemble`, `compact`). Based on user's goal, implement a basic RAG strategy by having the `ingest` hook add document embeddings to a vector store and the `assemble` hook retrieve context for the prompt.

FIELD OPERATIONS

Self-Healing RAG Pipeline

Build an agent that uses the `ContextEngine`'s `ingest` hook to continuously add documents to a vector store. Use the `assemble` hook to retrieve relevant chunks for the prompt, and the `afterTurn` hook to evaluate the model's response for accuracy, automatically correcting or updating the vector store if the information was outdated or incorrect.

Sub-Agent Specialist Spawner

Create a main 'dispatcher' agent. Using the `prepareSubagentSpawn` hook, the dispatcher analyzes an incoming task and spawns a specialized sub-agent with a dedicated, isolated context. For example, a 'Code Writer' sub-agent gets only relevant file contents, while a 'DB Query' sub-agent gets only the database schema, preventing context pollution.

STRATEGIC APPLICATIONS