BeeAI ​

BeeAI is an open-source agent framework backed by IBM Research. It focuses on reliability and transparency — structured memory tracks what the agent knows, tool validation ensures inputs match expected schemas before execution, and observability logs every reasoning step. MCP extends these validated agents with external tool access.

Be
BeeAI
IBM Research · GitHub
TRANSPORT
Streamable HTTP ✓
PLATFORM
TypeScript · Framework
MCP VIA
Tool Registration

Enterprise-Grade Reliability for Agent Tools ​

BeeAI validates every tool call before execution — checking parameter types, required fields, and value constraints against the tool schema. MCP tool schemas feed directly into this validation system, so hallucinated or malformed tool calls are caught before they reach your server.

Structured memory means the agent maintains a typed knowledge base throughout a conversation. When an MCP tool returns data, BeeAI stores it in memory with metadata, making it available for future reasoning steps without re-fetching.

Features:

  • Tool validation — schema-based input checking before every tool call
  • Structured memory — typed knowledge persistence across conversation steps
  • Observability — full reasoning traces with tool call logging
  • IBM-backed — active development by IBM Research
  • TypeScript — native TypeScript framework with strong typing
  • Serializable state — pause and resume agent sessions
  • Granite models — optimized for IBM Granite but works with any LLM

Integration ​

1. Create a Token ​

In Vinkius Cloud, go to your server → Connection Tokens → Create. Copy the URL.

2. Register MCP Tools ​

Add MCP tools to your BeeAI agent configuration. The framework validates all tool schemas at initialization.

3. Run with Validation ​

Every MCP tool call is validated against the schema before execution. Invalid calls are rejected with clear error messages.

FAQ ​

How does BeeAI validate MCP tool calls? BeeAI checks tool call parameters against the MCP tool schema before sending the request. Type mismatches, missing fields, and invalid values are caught at the framework level.

What is structured memory? A typed knowledge store that persists through the conversation. MCP tool results are stored with metadata for later reference without re-fetching.

Does BeeAI work with non-IBM models? Yes. Optimized for Granite but supports OpenAI, Anthropic, and other providers.

Is BeeAI free? Open-source under Apache 2.0.