Skip to content
← Tools / Skills
ToolMicrosoft/10 trustVerified
TaskWeaver logo

TaskWeaver

Code-first agent framework that turns data-analytics requests into executable Python.

Visit official site →

About

TaskWeaver is Microsoft's open-source "code-first" agent framework for seamlessly planning and executing data analytics tasks. Instead of relying on static tool plugins, it translates user requests into complete, stateful Python code snippets — naturally handling rich data structures like pandas DataFrames, numpy arrays, and tensors. A Planner agent creates high-level task roadmaps while a Code Interpreter agent generates and executes Python code, with both chat history and code-execution history (including in-memory data) preserved across turns.

It supports custom Python plugins with rich type annotations for enterprise databases, code verification before execution, Docker container-mode execution, shared memory between roles, and vision input for the Planner. The repo has around 6,100 GitHub stars and is MIT licensed. Honest limitations: it requires Python expertise to set up and extend, is focused on data analytics rather than general-purpose agency, and documentation can be sparse for advanced scenarios.

What it can do

  • Code-first planning: user requests become executable stateful Python snippets
  • Two-tier Planner/Code-Interpreter architecture with persistent runtime memory
  • Custom plugin authoring as Python functions with type annotations
  • Code verification before execution and containerized sandbox execution

Setup

pip install from the microsoft/TaskWeaver GitHub repo; all-in-one Docker image available for a one-stop experience.

Prerequisites

Python 3.9+, an LLM API key (supports OpenAI and local models), Docker optional for container-mode code execution.

Use cases

ai-agentsdata-analyticscode-generationpython

Works with

Trust score — 8/10

/10 trustVerified

Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 71; Trending score 76/100 from researcher assessment.

★ 6,162 GitHub stars

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.