nearai/ironclaw
Ironclaw
Ironclaw is an LLM orchestration framework and AI agent gateway designed to connect large language models with external tools, messaging interfaces, and persistent memory systems. It functions as a communication layer that routes interactions between users and AI models via HTTP webhooks and various messaging channels.
The system focuses on secure tool execution through a WebAssembly sandbox and isolated containers, which allows the framework to run untrusted code and dynamically generate new tools from natural language descriptions. Security middleware provides prompt injection defense and secret credential protection by injecting keys at the boundary and scanning outputs to prevent leaks.
The platform manages AI memory using a local encrypted database that combines full-text and vector search for context retrieval. It also includes an automated task scheduling system to handle recurring background jobs and maintenance routines via event triggers.
Features
- Agentic LLM Frameworks - Provides a platform for building autonomous agents with integrated tool use and persistent memory.
- Agent Connectivity Interfaces - Implements connectivity interfaces that bridge AI assistants to users via CLIs, web gateways, and messaging channels.
- AI Provider Integrations - Provides configuration interfaces to connect the framework to various external and local large language model providers.
- AI Gateways - Provides a communication layer that routes interactions between users and AI models via HTTP webhooks and messaging channels.
- Dynamic Tool Generation - The system generates new functional tools from natural language descriptions using WebAssembly to expand system capabilities.
- WebAssembly Tool Compilers - Compiles natural language descriptions into functional WebAssembly modules for safe, dynamic tool execution.
- Hybrid Search Engines - Combines keyword indexing and semantic embeddings to retrieve relevant context from local persistent storage.
- Vector Memory Stores - Implements a hybrid storage system combining vector similarity and keyword indexing for AI agent context retrieval.
- Containerized and Isolated Workspaces - Runs untrusted external code in isolated containers with defined resource limits to prevent host system compromise.
- Message Routing - Routes messages between diverse communication channels and core logic using a standardized internal message format.
- Encrypted Persistence - Stores user data and audit trails in a local database using strong encryption to ensure privacy.
- Local Data Storage - Uses strong encryption to store user information and audit logs in a local database for privacy.
- Secure Execution Environments - Runs untrusted code in isolated, resource-constrained environments to expand AI capabilities without compromising the host.
- Background Task Schedulers - Executes recurring background jobs using schedules, event triggers, and heartbeat monitors.
- Event-Driven Triggers - Triggers background maintenance routines and jobs using a system of heartbeats and timed event listeners.
- AI Security - Implements a protection layer to filter prompt injections and prevent credential leakage during model inference.
- Isolated Execution Sandboxes - Runs untrusted external code in isolated, resource-constrained containers to protect the host system.
- Wasm Sandboxes - Utilizes a WebAssembly sandbox to safely execute dynamic tools generated from natural language.
- Runtime Credential Injection - Securely injects encrypted credentials into the execution context at the boundary to prevent leakage in logs.
- Input Validation & Sanitization - Filters external prompt data through a detection layer to identify and block malicious injection attacks.
- LLM Prompt Injection Prevention - Filters external input using pattern detection and sanitization to prevent malicious prompt injection attacks.
- Credential Leak Prevention - Protects sensitive keys by injecting them at the execution boundary and scanning outputs to prevent accidental leaks.
- Credential Management Tools - Safely injects secret keys into workflows and scans outputs to prevent accidental credential leaks.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
