Flow Console

Overview

Flow Console is a desktop workflow engine that brings AI orchestration to your local machine. It connects to local LLM servers (via LM Studio or any OpenAI-compatible endpoint), runs semantic retrieval pipelines, and provides a clean interface for building multi-step AI flows without depending on external cloud services.

At its core, Flow Console is built around the Model Context Protocol — the same open standard that powers integrations across the Metantel product ecosystem. This lets Flow Console act as a unified control plane, sending commands and querying state across applications from a single desktop interface.

Core Capabilities

Local LLM connectivity — plug in any LM Studio model or OpenAI-compatible endpoint with zero configuration
Semantic retrieval — built-in vector search pipeline for RAG workflows across local document collections
Polyretrieval — hybrid local retrieval combining dense semantic search with sparse keyword recall for higher precision and coverage
Flow management — build, save, and replay multi-step AI task flows from a visual interface
Python client tools — scriptable automation layer for integrating Flow Console into custom pipelines

Local-First Retrieval Stack

Your Data Never Leaves the Machine

Flow Console runs its full retrieval pipeline locally — embedding, indexing, hybrid ranking, and re-ranking all happen on-device. Cloud LLMs become optional, not mandatory. The same retrieval surface powers the desktop UI and any remote MCP client connected to it, so the answers you see in Flow Console are the same answers an agent gets when it queries from elsewhere.

Remote Control via Elenchus

The Dispatch Protocol

Flow Console exposes its agentic capabilities over MCP, making them remotely accessible from the Elenchus mobile app. Through the Dispatch integration, users can dispatch agentic coding tasks, monitor running flows, and manage Flow Console sessions directly from Elenchus — without ever sitting at their desktop. Elenchus becomes the mobile command centre; Flow Console is the engine that executes.

Protocol-Native Architecture

Flow Console runs as a persistent control plane on the local machine, registering its agentic tools and flow state as callable resources. Connected clients — including Elenchus and the Artel connector — can invoke these tools over a single, consistent protocol, triggering and querying desktop AI tasks from any device.

This architecture means the heavy computation stays on the user's machine while control can travel anywhere. The Artel workspace connector follows the same pattern, demonstrating how the protocol scales across the Metantel product ecosystem.

Ecosystem Connections

Flow Console doesn't ship in isolation. It plugs into the rest of the Metantel stack through a shared protocol surface — context flows in and out without leaving the family.

Flow Console

Dispatch surface

Elenchus

Elenchus consumes Flow Console as a remote MCP server — your phone fires the command, the desktop executes.

Open Elenchus →

Flow Console

MCP target

Artel

Five live tools let Flow Console agents read and write inside Artel workspaces — docs, tasks, and decisions.

Open Artel →