Rerun: Turn your AI subscriptions into a team of coworkers
Rerun, listed on Product Hunt as launched on November 5, 2026, lets people run a group of AI coworkers on the ChatGPT or Claude plan they already pay for. Each coworker has its own tools, a computer, and access to a shared workspace, and it can run on a timer or in response to a trigger. The makers, Clément and Théo, describe each coworker as having a real browser that users can watch and take over. At the time of this brief, the Product Hunt listing showed 243 upvotes and 14 comments.
The new element is the open source release. Rerun's code is published under the AGPL-3.0 license, and users can clone the repository and start it with docker compose up on their own server. Coworkers can draw on a Claude or ChatGPT plan, any model reached through OpenRouter, or a local model served through Ollama, LM Studio, vLLM or another OpenAI-compatible server. The hosted Rerun Cloud runs the same code and is ready in about 30 seconds. Under AGPL-3.0, anyone who hosts a modified version for others must publish those changes under the same license.
A lead agent called the Foreman interviews the user and plans the team. Coworkers pause for approval before anything sensitive. The hosted plan starts at $29 a month for five agents, with unlimited seats, a private machine and unlimited runs, while the self-hosted version is listed as free.
Why it matters
For founders, the model shifts inference cost to the user. Buyers who already pay for a Claude or ChatGPT plan can avoid a separate model bill, so the vendor's economics rest on hosting, orchestration and software rather than tokens. That lowers the barrier to trying the product, but it also ties the product to the terms of the subscription it runs on.
For investors and operators, the AGPL choice is a distribution bet. Open code invites outside contributors and self-hosting in settings where data must stay in-house, while the hosted tier carries the revenue. Buyers evaluating agents will ask about approval flows and where data lives, so visible approval steps and self-hosting are selling points rather than features to add later.
Key points
- 1Each coworker gets its own tools, a real browser that users can watch or take over, memory and a schedule.
- 2The code is published under AGPL-3.0 and can be started on a personal server with docker compose up.
- 3The hosted plan starts at $29 a month for five agents, unlimited seats, a private machine and unlimited runs.
- 4Models can come from a Claude or ChatGPT plan, an API key or a local server such as Ollama.
- 5A lead agent called the Foreman interviews users and plans the team they need.
