BondAI is an open-source Python library (pip install bondai) for building and composing AI agent systems. It abstracts away common implementation complexities that developers face when building agents, including conversation memory and context window management, error handling and retries, and vector/semantic search for long-term memory storage. The framework supports multiple agent architectures out of the box: ReAct agents (reasoning and acting in a loop), Conversational agents for interactive dialogue, and Multi-Agent systems where agents can collaborate or delegate work to one another. A Tiered Memory system allows agents to store and retrieve information across sessions using vector embeddings. BondAI ships with a rich set of ready-made tool integrations. These include OpenAI models (GPT-4, GPT-3.5, DALL-E 3, Embeddings) and Microsoft Azure OpenAI, Google Search (via API key and CSE ID), Alpaca Markets for stock and crypto trading, PostgreSQL database querying with natural language, and Bland.ai for phone call automation and transcript processing. The framework also integrates with LangChain tooling. Deployment options include a CLI (run 'bondai' after install for a pre-configured interactive agent), a Docker image (krohling/bondai), or embedding the library directly into a custom Python codebase. The last published PyPI release was version 0.3.1 (January 2024), indicating the project has been in maintenance mode with no major new releases through mid-2026. Key features: - ReAct, Conversational, and Multi-Agent architecture support - Tiered memory system with vector/semantic search for long-term context - Built-in tool integrations: Google Search, PostgreSQL, Alpaca Markets, Bland.ai phone calls - OpenAI and Azure OpenAI model support including GPT-4 and DALL-E 3 - LangChain tool compatibility - CLI for zero-config interactive agent sessions - Docker image for containerized deployment - Automatic error handling and context window management
Free and open-source (MIT license). No commercial pricing tiers.