DISCOVER THE FUTURE OF AI AGENTSarrow_forward

Weaver

calendar_todayAdded Apr 24, 2026
categoryAgent & Tooling
codeOpen Source
PythonTypeScriptWorkflow AutomationNext.jsLangGraphFastAPIModel Context ProtocolMultimodalAI AgentsWeb ApplicationBrowser AutomationAgent & ToolingDeveloper Tools & CodingAutomation, Workflow & RPAEnterprise Applications & Office

An enterprise-grade AI Agent platform built on LangGraph, featuring Deep Research, code execution, browser automation, multimodal interaction, 62+ built-in tools, and intelligent routing.

Weaver is an enterprise-grade AI Agent platform with a decoupled frontend-backend architecture, using LangGraph as its workflow orchestration core. An intelligent router automatically selects among four strategies: direct (straightforward answers), web (web search), agent (single-step tool calls), and deep (deep research).

The platform includes 62+ built-in tools covering E2B sandbox code execution, Playwright browser automation, file and Office document generation, and optional desktop control. It also supports third-party tool extension via MCP (Model Context Protocol). The Deep Research mode features parallel retrieval, multi-round iteration, and clickable citation sourcing within the UI.

The backend is built on FastAPI + LangGraph, providing SSE streaming output and OpenAPI documentation. The frontend uses Next.js for streaming rendering, tool activity panels, and virtual scrolling for long code outputs. Frontend-backend contract alignment is achieved through automatic OpenAPI-to-TypeScript type generation. It is compatible with multiple LLM backends including OpenAI, Anthropic Claude, and DeepSeek.

Deployment supports local development (Python 3.11+ / Node.js 18+) and Docker Compose (with optional PostgreSQL/Redis). Core environment variables include OPENAI_API_KEY, TAVILY_API_KEY, E2B_API_KEY, etc. Latest version v0.6.7, with 586 commits and 9 releases. Primary languages are Python (83.7%) and TypeScript (15.2%).

Use Cases: Deep research and knowledge synthesis, programming assistance and sandbox code execution, browser automation and data scraping, enterprise-grade complex workflow orchestration, document and Office file processing.

Boundaries: Does not provide self-hosted LLM inference; relies on external APIs. No built-in model training or fine-tuning pipelines.

Unconfirmed Information: Online demo URL and standalone documentation site not extracted; full list of 62+ tools requires checking the tools/ directory; Manus-like multi-Agent configuration mechanism not detailed; MIT License notes Copyright (c) 2026, which may be a typo.

Related Projects

View All arrow_forward

STAY UPDATED

Get the latest AI tools and trends delivered straight to your inbox. No spam, just intelligence.

rocket_launch