I built BHive: a multi-agent framework for Gemini/OpenAI/Anthropic/Ollama. Tear it apart.

Description and use

BHive is a TypeScript multi‑agent framework that abstracts over Gemini, OpenAI, Anthropic, and Ollama, enabling task delegation and tool calling with a single codebase.

Problem

Developers need a flexible, provider-agnostic framework to orchestrate multi-agent workflows with tool calling and delegation across various LLM providers.

Outcome

BHive provides a unified, extensible framework for multi‑agent LLM workflows, enabling seamless provider switching and task delegation while maintaining high test coverage.

Workflow

• Define LLMAdapter interface to abstract provider APIs.
• Implement ToolCallingParser to normalize tool‑call formats.
• Create Bee agents that can delegate tasks using createDelegationTool.
• Validate tool calls in Bee execution loop with ToolCallValidator.
• Configure timeout resolution via explicit args, env vars, or defaults.
• Run tests to ensure 98% coverage.
• Seek community feedback on security model, delegation abstraction, and repo structure.

Evidence

BHive is a multi-agent framework in TypeScript. Agents (“Bees”) can delegate tasks to each other, call tools, and it works with any LLM provider — Gemini, OpenAI, Anthropic, or local Ollama — through the same code.

Source