Octop Harness

Overview

Octop Harness is a production-grade agent runtime based on the Harness Engineering philosophy and an engineering wrapper around deepagents. It extends the create_deep_agent primitive into a production-ready runtime, offering multi-provider model routing, persistent tiered memory, browser and search tools, a multi-agent registry, pluggable storage backends, and a terminal CLI. Meanwhile, Octop serves as a self-hosted assistant that provides parallel-operating intelligent collaboration for individuals and teams.

Key Features

• Built on deepagents' create_deep_agent, it provides a production-oriented agent encapsulation and runtime. ChatModelFactory supports OpenAI, Anthropic, AWS Bedrock, and 17 editable provider presets. Tiered memory captures each turn as L0 raw events and asynchronously distills

Use Cases

• Individuals and teams can deploy a service on one machine to provide separate dedicated assistants for family and colleagues, switching roles such as analyst or researcher as needed. After authorizing connections to Tencent Docs, Tencent Meeting, Notion, email, and other

Getting Started

• Prerequisites: Python 3.12+ and a model provider API key (OpenAI / Anthropic / Bedrock / compatible). Install the core SDK with pip install octop-harness; for the terminal CLI, use pip install octop-harness[cli]. Initialize and configure a provider by running octop-harness init

Deployment & Requirements

• Requires Python 3.12+. Requires a model provider API key (OpenAI / Anthropic / Bedrock / compatible). Optional dependency extras include cli, bedrock, object-storage, desktop, web-search-all, remote-backends, observability, acp, and all; missing extras fail at use-time with an

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.