How do I start a new review project with Weave?

Simply use the push_plan tool and omit the document_id. The server will automatically create a new project and document for your current task.

Does Weave support multi-agent or team workflows?

Yes. While subagents cannot call MCP tools directly, they can send their content to a lead agent who then pushes the combined work to Weave for review.

What is the primary purpose of the Weave skill?

Weave acts as a review bridge, allowing Claude to submit plans and code changes to humans via a web UI to ensure accuracy and alignment before proceeding with tasks.

Can Weave be used for non-coding tasks?

Yes. Weave is a general-purpose plan review tool. It can be used for business plans, budgets, roadmaps, and architecture docs, not just software engineering tasks.

How does Claude handle feedback from a human reviewer?

Claude uses the pull_plan or pull_code_feedback tools to retrieve specific comments and decisions, allowing it to address issues and re-push updated versions.

Weave AI Review Bridge

Name: Weave AI Review Bridge
Author: weave-ai-dev

byweave-ai-dev

•

协作工具

Bridges AI agents and humans through a dedicated review interface for plans, code diffs, and structured feedback.

Weave is an MCP-powered skill that establishes a collaborative review bridge between AI agents and human developers. It enables Claude to push proposed plans, architectural designs, and code diffs to a rich web editor where humans can comment, request changes, or grant approval. By providing a structured way to pull line-specific feedback back into the terminal session, Weave facilitates an iterative, human-in-the-loop workflow that ensures AI-generated output remains high-quality and aligned with project requirements. It is particularly useful for complex tasks that require human oversight, such as major refactors or initial project architecture.

主要功能

01Bi-directional feedback loop to pull human comments directly into Claude

02Project-based grouping and document versioning for iterative development

033 GitHub stars

04Side-by-side code review interface with Monaco-based diffing

05Seamless push of architectural and project plans for human sign-off

06Automated OAuth authentication for secure remote MCP communication

使用场景

01Design and Planning: Getting human approval on feature roadmaps before implementation begins

02Code Quality Assurance: Submitting significant code changes for a formal peer review by a human

03Multi-agent Coordination: Using a lead agent to aggregate and push team proposals to a human manager

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add weave-ai-dev/agent-skills weave

For use in Claude.ai and ChatGPT