Ratel

Ratel builds an open-source context gateway for developers creating AI agents. Its TypeScript and Python SDKs share a Rust engine that retrieves relevant tools and skills for each turn instead of placing the entire catalog in the prompt. Developers can search and run tools, MCP servers, and skills through a common interface, then inspect retrieval and invocation through local traces or OpenTelemetry. Ratel Local brings the same capabilities to Claude Code, Codex, and Cursor without requiring SDK integration.

The company was cofounded by CEO Giacomo Nicoli and CTO Roberto Stagi. Its focus emerged from Agentified, an earlier effort to embed agents in SaaS applications, after the founders encountered the difficulty of supplying useful model inputs. Ratel uses in-process BM25 keyword search without a vector database or API key, with optional semantic and hybrid ranking. Experimental adaptive ranking also uses search-and-invocation history to influence future capability selection.

Ratel distributes its engine under Apache-2.0 and its SDKs and other components under MIT, permitting commercial use, modification, and redistribution. This lets developers embed the gateway in their own applications. Its planned Ratel Cloud service extends that approach to hosted analytics for context usage and retrieval performance across agents; it remains coming soon.