BabyAGI
The 2023 task-loop agent, now a family of open-source experiments
Overview
BabyAGI is a set of open-source agent experiments by Yohei Nakajima. The March 2023 task-creation loop is archived; the main repository now holds functionz, an experimental framework for self-building agents, and a separate BabyAGI 3 repository (2026) offers a minimal personal agent with memory and scheduling. All of it is free and explicitly not production software.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
What you can produce with BabyAGI
- The original March 2023 task-creation, execution and prioritisation loop, preserved in the babyagi_archive repository
- functionz, a framework that stores functions in a database and tracks their dependencies as a graph
- A local dashboard for managing functions, running updates and reading execution logs
- Two experimental self-building agents that write new functions from existing ones
- BabyAGI 3, a minimal personal agent with persistent memory, scheduled tasks and web search
- A compact codebase for teaching how agent loops behave and where they fail
BabyAGI matters as history and as teaching material. The archived 2023 loop is still the clearest short illustration of an agent creating, executing and re-prioritising its own tasks, and the current functionz framework and the newer BabyAGI 3 repository show where Yohei Nakajima took the idea next. The main repository has not changed its code since late 2024, an automated readiness report scores it 3 out of 10, and its README says plainly that it is not for production.
Use it to learn, prototype and borrow ideas. For anything that has to run reliably, choose a maintained agent framework. It runs locally on your own API keys, so there is no regional access issue.
What we found
BabyAGI is the name Yohei Nakajima has given to a series of small experiments in autonomous agents, and the main GitHub repository no longer holds the script that made it famous. The original, released in March 2023, was a short Python loop in which a language model created tasks toward an objective, executed them, generated new tasks from the results and re-prioritised the queue, keeping results in a vector database. It spread alongside AutoGPT as one of the first widely copied agent patterns. That code was frozen in September 2024 in the babyagi_archive repository, under MIT.
The self-building framework now in the main repository arrived the same month. Its core, functionz, stores functions in a database, tracks their imports and dependencies as a graph, loads them automatically and logs every run, with a web dashboard and two experimental agents that write new functions from existing ones. The README is candid: Nakajima has never held a job as a developer, the code exists to share ideas, and it is not meant for production. The last PyPI release, 0.1.4, dates from October 2024. The only 2026 change is an automated code-readiness report added in January, which scores the framework 3 out of 10 and flags running stored code through exec() with no sandbox as a critical risk. The repository has no licence file, although its README says MIT.
Active work has moved to sibling repositories. babyagi-2o (October 2024) chases the smallest possible self-building agent, and babyagi3, started in January 2026 under MIT and last updated in March, is a personal assistant you configure once and then run in plain language, with persistent memory, scheduled tasks, web search and optional email and SMS channels through third-party services. Its README warns about running costs and about exposing its API beyond localhost.
Everything here is free and runs on your own model API keys. Read it to understand how agent loops work and fail; build production systems on maintained frameworks.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
Notable facts
- BabyAGI was created by a venture capitalist (Yohei Nakajima) as a weekend experiment to test whether LLMs could autonomously pursue goals — it became one of the most-starred AI repositories in GitHub history.
- The entire BabyAGI codebase is under 200 lines of Python — an extraordinarily simple implementation for a concept that influenced much larger frameworks.
- BabyAGI's academic paper on task-driven autonomous agents has been cited in foundational research on LLM agent architectures.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including BabyAGI's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with BabyAGI unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to BabyAGI directly →
Spotted something out of date? Suggest an update →
Related reading
Health & Wellness
Fine-Particle Air Pollution Linked to Higher Hospital Admission Risk for Up to 21 Days, St...
Cybersecurity
SAP Rates a Python SDK Flaw 10 Out of 10. Whether It Hits You Depends on One Field
Health & Wellness
A Guide for New Mothers Missed Its Main Goal but Nearly Doubled Primary Care Visits
Alternatives to BabyAGI
More in Agents & Automation
