ericPavone/openClaws

OpenClaws: Enhanced OpenClaw Fork with Claude Code and MongoDB Integration

OpenClaws is a sophisticated fork of the rogerli007/my-openclaw-page framework that introduces native support for Claude Code as an alternative brain backend and MongoDB as a persistent knowledge base. Developed by ericPavone, this project addresses critical limitations in the upstream OpenClaw implementation, particularly around CLI agent capabilities and memory persistence.

The project is in active development, with its most recent update pushed on February 6, 2026. While currently showing zero stars and forks, openClaws represents a significant architectural enhancement for developers seeking more robust AI agent orchestration with enterprise-grade persistence layers.

Key Features and Improvements

Claude Code Native Integration

The most significant improvement in openClaws is the removal of tool restrictions for Claude Code integration. The upstream OpenClaw implementation injects "Tools are disabled in this session" into the system prompt, preventing Claude Code from utilizing its native capabilities like Read, Write, and Bash tools. OpenClaws introduces an extraSystemPromptOverride configuration in CliBackendConfig that allows Claude Code to operate without restrictions while maintaining backward compatibility with other CLI backends.

This modification touches multiple core files including cli-runner.ts, cli-backends.ts, and configuration type definitions, enabling Claude Code to function as a first-class brain option alongside traditional API-based LLM providers.

Hook Injection for CLI Agents

Another critical enhancement addresses the disparity between CLI and API agents. In the original OpenClaw, plugin hooks like before_agent_start only function with API agents, leaving CLI agents without access to plugin context such as auto-recall, configuration injection, and memory systems.

OpenClaws extends hook functionality to CLI agents by invoking before_agent_start in the CLI runner, ensuring that Claude Code and other CLI-based agents receive the same contextual information as their API counterparts. This includes memories, configuration data, and persona definitions that are critical for maintaining consistent agent behavior.

MongoDB Memory Plugin

The centerpiece of openClaws is the memory-mongodb plugin, a native TypeScript implementation that replaces the deprecated Python-based mongoBrain skill. This plugin manages five specialized MongoDB collections:

• memories: Facts, preferences, decisions, and entities with TTL-based auto-capture
• guidelines: Behavioral rules with priority and domain specification
• seeds: Reusable knowledge templates
• agent_config: Per-agent configuration including soul, identity, persona, and instructions
• skills: Skill definitions with trigger matching capabilities

The plugin provides auto-recall functionality that injects relevant memories into agent context based on prompt analysis, auto-capture for automatically detecting and storing memorable statements, and agent config injection that loads personality and behavioral parameters at runtime.

Installation and Configuration

Setting up openClaws requires configuring Claude Code as the primary brain and enabling MongoDB persistence. For a Claude Code-only configuration, users specify claude-cli/opus as the primary model in the agents configuration. The system requires the Claude CLI to be installed and authenticated via claude login, eliminating the need for API keys.

For multi-model deployments, openClaws supports fallback configurations where primary models like google/gemini-3-pro-preview can fall back to Claude Code when needed. The MongoDB plugin configuration requires a MongoDB URI, database name, agent ID, and flags for enabling auto-recall and auto-capture features.

The project includes workspace bootstrap functionality that automatically appends mongoBrain sections to TOOLS.md and BOOT.md when CLI models are configured, streamlining the setup process.

Technical Architecture

OpenClaws implements a sophisticated message routing flow: incoming messages from channels (WhatsApp, Telegram, Discord) pass through the gateway to routing bindings, which determine whether to use CLI or embedded API agents. Before agent execution, the before_agent_start hooks inject agent configuration and relevant memories, then either spawn a Claude CLI subprocess or make API calls to providers like Anthropic, OpenAI, or Google AI.

The project structure uses Git submodules, with the core OpenClaw fork in the opencliclaw/ directory and the deprecated Python mongoBrain in a separate directory for reference.

Deployment and Operations

The repository includes production deployment configurations for EC2 environments using systemd service management. The installation leverages nvm for Node.js version management, with OpenClaw installed at version 24.13.0. Command-line tools are provided for MongoDB operations including status checks, search, storage, configuration retrieval, skill matching, export, pruning, and migration from workspace files.

Community and Development Status

As a newly published fork, openClaws currently has no stars, forks, or open issues. The repository lacks a license specification, which may impact adoption for commercial use cases. The project has no assigned topics or description, suggesting it's in early public release stages. The codebase is written primarily in TypeScript with no dominant language detected yet, and it is not archived, indicating active development intent.

Comparison with Alternatives

Compared to upstream OpenClaw, openClaws offers superior CLI agent integration and enterprise-grade persistence. While standard LangChain or AutoGPT implementations require custom memory solutions, openClaws provides batteries-included MongoDB persistence with semantic search and auto-capture capabilities. The Claude Code integration offers a unique advantage for developers seeking local-first AI agent architectures without API dependencies.

For teams requiring multi-channel AI agent orchestration with persistent memory across conversations, openClaws presents a compelling alternative to building custom solutions from scratch, particularly in Italian-language contexts where the documentation is primarily provided.