RadarTopicsBuildersWeeklyReads
Open Source Radar
PrefectHQ/

marvin

GitHubWebsite

Marvin is a Python framework for producing structured outputs and building agentic AI workflows. It provides tasks, agents, threads, and planning/orchestration to manage multi-step AI processes.

6.2kstars
407forks
110issues
Apache-2.0license
2023since
Star historydaily snapshots by VibeCrowd

Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).

Alternatives & relatedmatched by topic overlap
Reviewgenerated from repository data · Aug 5, 2026

What it is

Marvin is a Python framework for producing structured outputs and building agentic AI workflows. It provides an intuitive API for defining workflows and delegating work to LLMs, including functions to cast, classify, extract, and generate structured data, plus the ability to create discrete tasks and assign specialized agents.

How it works

Marvin centers on tasks, agents, and threads to orchestrate AI work:

  • Tasks describe objectives with observable outputs and can depend on other tasks.
  • Agents are specialized AI configurations that can be assigned to tasks.
  • Threads manage context sharing and progress across multiple tasks.
  • There are high-level helpers like marvin.run, marvin.summarize, marvin.classify, marvin.extract, marvin.cast, and marvin.generate for common patterns.

Getting started

Installation is described in the README and package on PyPI. Example installation commands shown:

uv add marvin

To configure LLMs, the README indicates setting environment variables, for example:

export OPENAI_API_KEY=your-api-key

The repository provides usage examples across structured-output utilities, agent usage, and planning/orchestration.

Recent releases

Latest releases include:

  • v3.2.7 (2026-03-04): fix(mcp) merges custom env with os.environ for MCPServerStdio, addressing environment propagation in subprocesses.
  • v3.2.6 (2026-01-22): OpenAI Observability Integration (Beta) for tracing Marvin agent calls to OpenAI observability dashboards.
  • v3.2.5 (2026-01-06): MCP Server Lifecycle Fix—prevents restarting MCP servers for every agent.run() call.
  • v3.2.4 (2025-12-22): Docs fixes removing stale delegates parameter from Agent documentation.
  • v3.2.3 (2025-11-03): Other changes including using Claude Haiku 4.5 for research agent.

Traction

Stars: 6184. Forks: 407. Open issues: 110. Language: Python. License: Apache-2.0. Created: 2023-03-10. Last push: 2026-07-31.

Behind the repo

Not provided in the content.

Caveats

License: Apache-2.0. Created: 2023-03-10. Last push: 2026-07-31. Open issues: 110. No migration notes beyond the release notes; database migration details are not included in the excerpt.

SharePost on XLinkedIn
All trending reposRevenue-verified startups →