Reworkd archived the AgentGPT GitHub repository on January 28, 2026 — a "This repository was archived by the owner" banner now sits at the top of the project page, and it's read-only, meaning no more updates, fixes, or new features are coming. The hosted site at agentgpt.reworkd.ai still loads and still runs in a limited free state, so you technically can still spin up an autonomous task agent today. Whether that's worth doing is a different question, and this is the honest answer.
What AgentGPT actually was
AgentGPT launched in 2023 as one of the first no-code, browser-based tools for building autonomous AI agents — set a goal in plain language, and the agent would think, plan, and work through sub-tasks on its own using GPT-3.5 or GPT-4. It became one of the breakout demos of the early "AutoGPT" wave, with sample agents like TravelGPT for trip planning and ResearchGPT for report drafting showing off what an unsupervised goal-driven agent could attempt. At its peak, the appeal was real: zero setup, watch an agent reason through a task live in your browser.
The pricing before archival was freemium, with a Pro tier around $40/month unlocking GPT-4, unlimited web search, and the latest plugins on top of the free GPT-3.5-based tier.
Why it stalled
The core problem was never the idea — it was reliability. Independent reviews consistently describe the same pattern: agents that loop on the same sub-task without resolving it, output that stays shallow relative to the goal given, and runs that burn through usage credits without ever reaching a finished result. That's a hard limitation to fix with an unmaintained codebase, and it's likely part of why Reworkd — the small team behind it — moved on rather than continuing to invest engineering time in a category that more capable, better-funded agents (Manus, Devin, Claude Code) had by 2026 mostly outgrown.
What still works today, and what doesn't
The hosted demo at agentgpt.reworkd.ai reportedly still loads and lets you run a basic agent in the free tier — useful only as a historical look at what the early autonomous-agent wave actually looked like in practice. What you should not do: rely on it for real work, expect the Pro tier or GPT-4 access to function reliably, or plan around any of it getting fixed. An archived repository gets no security patches, no API updates when the underlying model providers change their endpoints, and no bug fixes — it will only get less reliable from here, not more.
What to use instead
[Manus](/go/manus) is the closest modern equivalent to what AgentGPT was reaching for — a general-purpose autonomous agent for research, planning, and multi-step web work, actively developed and freemium to start. [Relevance AI](/go/relevance-ai) is the better pick if you want to build a repeatable custom agent for a specific business task rather than run one-off experimental goals — it's a no-code builder aimed at handling actual recurring work, not demos. Neither one recreates AgentGPT's exact "type a goal, watch it think" simplicity, but both are maintained, and that matters more in 2026 than it did during the 2023 hype cycle.
FAQ
Is AgentGPT safe to use even though it's archived? The hosted demo itself doesn't appear to be malicious, but treat any account or API key you connect to it with the same caution you'd apply to any unmaintained service — don't give it access to anything sensitive, since there's no one left to patch a security issue if one surfaces.
Did Reworkd shut down as a company? Archiving a specific GitHub repository means that project is no longer maintained; it doesn't necessarily mean the team itself has closed. Check Reworkd's own channels directly for their current status and any other products they may be working on.
Can I still self-host the old AgentGPT code? The archived repository is read-only but still publicly viewable, so the code remains available to clone and self-host. You'd be responsible for your own maintenance, model-API compatibility, and security going forward, since Reworkd itself will not be updating it.
Is there a direct, actively maintained "AutoGPT-style" alternative? Manus is the most direct spiritual successor in terms of user experience — describe a goal, let the agent work autonomously — and it's under active development rather than archived.
Related guides
- Best Autonomous AI Agents in 2026: Manus vs Devin vs Replit Agent
- Best No-Code AI Agent Builder in 2026
- AI Agents vs. Automation: What's the Difference?
- Explore Agents & Automation tools
*News and pricing details reviewed as they change. Last updated September 2026.*
Bogdex · Founder & editor, woska
Bogdex builds and curates woska, testing AI tools against real workflows to judge which ones actually save time rather than which have the longest feature list.