BondAI

What is BondAI?

BondAI is an open-source Python framework for developing single-agent and multi-agent systems. It handles common implementation work such as context and memory management, error handling, vector search, semantic retrieval, tool execution, and communication between agents. The framework supports ReAct-style agents, conversational agents, and multi-agent patterns.

Its tiered memory design separates core memory, complete conversation history, and archival memory. Core memory keeps essential current information in the system prompt, conversation memory provides keyword search across prior dialogue, and archival memory uses semantic search for larger collections. A MemoryManager gives agents tools for accessing and editing these stores and updates the prompt with relevant information.

BondAI includes built-in tools for Python, shell commands, files, websites, downloads, databases, search, email, phone calls, and task delegation. Developers can create custom tools or import LangChain tools. It also provides a command-line interface and REST or WebSocket agent server.

How to Use BondAI

1. Install and configure the BondAI package and the required model-provider credentials.
2. Choose an agent type, such as a basic, conversational, ReAct, or multi-agent configuration.
3. Add only the tools required for the workflow.
4. Configure core, conversation, and archival memory sources when persistent context is needed.
5. Run the agent directly, through the CLI, or through the REST or WebSocket server.
6. Place code-execution and shell tools in an isolated Docker environment before allowing an agent to use them.

Core Features

• Single- and multi-agent patterns: Supports ReAct, conversational, and delegated agent systems.
• Tiered memory: Manages core, conversation, and archival information with different retrieval methods.
• Built-in tools: Includes code, shell, file, website, database, search, email, and other integrations.
• Custom tool support: Lets developers implement tools or import compatible LangChain tools.
• Semantic retrieval: Queries large files, websites, and archival memory through embeddings.
• Multiple interfaces: Provides Python APIs, CLI access, and REST or WebSocket serving.

Use Cases

• Research agents: Combine web search, website querying, files, and long-term memory.
• Data assistants: Connect PostgreSQL and answer natural-language questions about a schema.
• Multi-agent workflows: Delegate specialized subtasks through an agent hierarchy.
• Automation agents: Use APIs, email, phone, or controlled shell tools to complete multi-step tasks.
• Embedded agent services: Expose a configured agent to an application through REST or WebSocket.

Frequently Asked Questions

Is BondAI open source?

Yes. The official introduction describes BondAI as an open-source agent-development tool.

Can BondAI use custom tools?

Yes. Developers can create tools or import them from LangChain.

Should shell and Python tools run directly on a host machine?

The documentation recommends using Docker because arbitrary generated commands can damage the host.