Monkeys Core Concepts

Monkeys Core Concepts

This page introduces the concepts used across the current Monkeys ecosystem.

Workflow

A workflow is the executable definition of an AI application. It describes how input data moves through model calls, tool invocations, control nodes, human tasks, and output construction.

Task

A task is a node inside a workflow. Tasks can call LLMs, invoke tools, transform data, branch execution, run loops, or wait for human input.

Conductor

Conductor is the workflow orchestration engine used by Monkeys. monkeys-server translates platform workflows into Conductor execution, while worker services execute long-running or external tasks.

Core Product

The core product is the minimum useful Monkeys platform:

• monkeys-server: main backend, workflow API, marketplace, auth, assets, tools, model configuration, and service bus.
• monkeys-studio: main user-facing Studio frontend.
• monkeys-kernel: platform and kernel management frontend.
• monkeys-compute: compute control-plane frontend.

Capability Services

Capability services extend the core product. Examples include monkeys-agent-server for agent runtime, monkeys-billing-server for billing, monkeys-data-server for data indexing, monkeys-webhook-server for notifications, and monkeys-mcp-server for MCP access.

Tools

Tools are HTTP services that expose a Monkeys manifest.json and an OpenAPI schema. A tool can be hosted inside Monkeys Server or deployed separately. Tool services include the sandbox, third-party API adapters, Python tools, and agent toolkits.

Assets and Data

Assets include media files, model assets, text data, table data, ComfyUI assets, design data, and other application resources. Data services keep PostgreSQL as the source of truth and use Elasticsearch for search acceleration where needed.

Agent

An Agent is a runtime entity with model configuration, instructions, tools, threads, messages, and tool-call history. Agent features are backed by monkeys-agent-server and can be connected through Studio, APIs, or MCP.

Integration

Monkeys can expose workflows through OpenAI-compatible APIs and can be operated through MCP. This allows chat clients, API gateways, AI coding tools, internal services, and custom clients to call or manage Monkeys capabilities.