OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Ratel

Ratel is an open-source context-engineering SDK and retrieval layer for AI agents. It indexes tools and skills in catalogs so an agent can search for relevant capabilities, invoke tools by ID and load skill instructions when needed. BM25 retrieval is the default; semantic and hybrid ranking are optional, and the project describes semantic retrieval using an in-process model or an OpenAI-compatible embedding endpoint.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Ratel — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Ratel when an agent can access more tools, skills or instructions than it needs for a typical turn, and you want to retrieve capabilities on demand rather than load the entire set up front.

Who it is for

  • Developers building AI agents with large or changing sets of tools and skills.
  • Teams integrating context retrieval into TypeScript or Python agent applications.
  • Builders who want an MCP-based route for exposing context-management functions.

Documented uses

  • Search tool and skill catalogs for capabilities relevant to an agent turn.
  • Invoke a selected tool by its catalog ID.
  • Load relevant skill instructions only when an agent needs them.
  • Retrieve agent memory using BM25 or optional semantic and hybrid ranking.
  • Reduce irrelevant tool schemas and skill instructions in agent context.
Before you choose

Requirements & limits

The repository is public and licensed under MIT.

  • Ratel is a developer integration layer; the documented examples connect its catalog functions to an agent framework rather than providing a standalone agent.
  • BM25 is the default; semantic and hybrid ranking are opt-in.
  • Semantic indexing and search are asynchronous, and the documented embedding choices are an in-process model or an OpenAI-compatible embedding endpoint.
  • The project states that it requires no vector database.
  • The repository identifies an MCP server capability in addition to its TypeScript and Python SDKs.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Ratel catalogs in a TypeScript agent

Install the SDK, register a tool and skill, and obtain the catalog functions to connect to an agent framework.

Before you start

  • A TypeScript project using pnpm.
  • A tool implementation and skill instructions to register.

Steps

  1. Install the package with pnpm add @ratel-ai/sdk.

  2. Create a ToolCatalog and register tools with their identifiers, descriptions, schemas and execution functions.

  3. Create a SkillCatalog and register skills, including their descriptions, associated tool IDs and instruction bodies.

  4. Create the search, invoke and skill-content functions with searchCapabilitiesTool, invokeToolTool and getSkillContentTool for use as agent tools.

Expected outcome

The project has registered tool and skill catalogs and the functions needed to connect catalog search, tool invocation and skill loading to an agent framework.

Guide limitations

  • The example documents catalog setup and the functions to connect; it does not describe completing or deploying an agent application.
  • The example does not specify a particular agent framework configuration.
Documentation 1 ↗
Source-based practical guide

Set up Ratel catalogs in a Python agent

Install the Python package, register a tool and skill, and obtain the catalog functions to connect to an agent framework.

Before you start

  • A Python project with pip.
  • A tool implementation and skill instructions to register.

Steps

  1. Install the package with the command below.

    Documented command
    pip install ratel-ai

  2. Create a ToolCatalog and register an ExecutableTool with its identifier, description, input schema and execution function.

  3. Create a SkillCatalog and register a skill with its description, associated tool IDs and instruction body.

  4. Create the search, invoke and skill-content functions with search_capabilities_tool, invoke_tool_tool and get_skill_content_tool for use as agent tools.

Expected outcome

The project has registered tool and skill catalogs and the functions needed to connect catalog search, tool invocation and skill loading to an agent framework.

Guide limitations

  • The example documents catalog setup and the functions to connect; it does not describe completing or deploying an agent application.
  • The example does not specify a particular agent framework configuration.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
9 October 2026
Original website title
ratel-ai/ratel · Official homepage title · retained in its source language
How we review sourcesSuggest a correction