Mini-Harness
A lightweight AI agent framework developed independently, with a UI style inspired by Claude Code. It supports multiple API integration methods (OpenAI/Anthropic formats) and integrates core modules including browser control, terminal access, file read/write, web search, sub-agents, and skills, emphasizing high integration, fast execution, and full controllability.
Timeline (last 90 days)
The author built a lightweight agent framework called Mini-Harness from scratch because off-the-shelf frameworks come with many unused modules that consume resources and slow down execution.
Mini-Harness takes visual inspiration from and pays homage to Claude Code in its interface style.
Mini-Harness supports multiple API integration methods, including the OpenAI legacy format, the OpenAI Response format, and the Anthropic-style interface, which users can select in the settings.
Mini-Harness integrates modules including browser control, terminal, file read/write, file upload, web search, sub-agents, and Skills.
The author demonstrated a case in which an agent autonomously completed a 3D subway simulation scene in a Counter-Strike style, with the entire process taking approximately 709 minutes (close to 12 hours).
Mini-Harness supports two workspace modes: local projects and remote SSH connections, allowing users to load a local directory or enter a username and password to connect to a remote SSH server.
Mini-Harness's search service is an optional configuration; the author uses SearXNG to set up a private search server, and when left unconfigured, the system uses a built-in method to retrieve content from the web.
Mini-Harness controls the local or specified browser in privileged mode to enable real web operation capabilities.
Mini-Harness's task area supports custom task types, such as the "brainstorming" mode customized by the author.
Mini-Harness can drive an agent to operate Word and PowerPoint, and can open Blender for 3D modeling and iteratively refine the output autonomously.
1 more timeline events
All Facts (11)
When a team has clear requirements for performance, controllability, and future scalability, building a lean harness from scratch is often more efficient than using a heavy-duty general-purpose framework.
50%UnverifiedMini-Harness takes visual inspiration from and pays homage to Claude Code in its interface style.
50%UnverifiedMini-Harness supports multiple API integration methods, including the OpenAI legacy format, the OpenAI Response format, and the Anthropic-style interface, which users can select in the settings.
50%UnverifiedMini-Harness integrates modules including browser control, terminal, file read/write, file upload, web search, sub-agents, and Skills.
50%UnverifiedThe author built a lightweight agent framework called Mini-Harness from scratch because off-the-shelf frameworks come with many unused modules that consume resources and slow down execution.
50%UnverifiedMini-Harness supports two workspace modes: local projects and remote SSH connections, allowing users to load a local directory or enter a username and password to connect to a remote SSH server.
50%UnverifiedMini-Harness's search service is an optional configuration; the author uses SearXNG to set up a private search server, and when left unconfigured, the system uses a built-in method to retrieve content from the web.
50%UnverifiedMini-Harness controls the local or specified browser in privileged mode to enable real web operation capabilities.
50%UnverifiedMini-Harness's task area supports custom task types, such as the "brainstorming" mode customized by the author.
50%UnverifiedMini-Harness can drive an agent to operate Word and PowerPoint, and can open Blender for 3D modeling and iteratively refine the output autonomously.
50%UnverifiedThe author demonstrated a case in which an agent autonomously completed a 3D subway simulation scene in a Counter-Strike style, with the entire process taking approximately 709 minutes (close to 12 hours).
50%