Reunit SA · Rerun: Open-Source AI Coworkers on Schedules, With Their Own Cloud Computer
Rerun is a no-code platform for building teams of AI coworkers that work in your apps on a schedule, on a webhook or when you message them, on a private cloud machine with a browser and a screen, and that can stop and ask for your approval before they send, pay or delete anything. It suits founders, small businesses and agencies that want recurring business work handed off, using the Claude, ChatGPT or OpenCode plan they already pay for, Rerun's included credits, or their own API keys, and it can be self-hosted under AGPL-3.0.
Best for
Founders, small businesses and agencies that want several always-on AI coworkers handling recurring business tasks across their apps, with approvals before consequential actions, flat pricing per pack of coworkers rather than per seat, and the option to run on a Claude or ChatGPT plan they already pay for or to self-host
Not ideal for
Teams that need a single deterministic workflow engine with fixed steps, organizations that need per-coworker isolation or fine-grained API permissions inside one workspace, buyers who want a free hosted tier, and anyone unwilling to run Docker and Supabase who still needs to keep data off a vendor's cloud
Who it's for
Founders, small businesses, freelancers and agencies that want recurring business work (inbox, sales prospecting, invoicing, social media, reporting) done by AI coworkers, plus technical teams that want to self-host the same product
Rerun's appeal is that it removes two common objections to hosted AI employees at once. Model cost can ride on a Claude, ChatGPT or OpenCode plan you already pay for, and lock-in is reduced by an AGPL-3.0 codebase you can run yourself. The design choices behind it are sensible for business work: an always-on machine per workspace, approvals that pause even 3am scheduled runs, secrets kept out of the model's context, and pricing per coworker rather than per seat or per run. The documentation is also frank about the gaps. Self-hosting is real but not complete, because the template library, paid data, LinkedIn and WhatsApp apps and email notifications stay on Rerun Cloud, and Google and Microsoft apps need your own Composio setup. Inside a workspace, coworkers share one machine and database, and they may edit their own prompts and install tools unless you switch that off. It suits small teams and agencies that want several coworkers on recurring tasks and are willing to review approvals. Teams that need strict isolation, mature governance or a long vendor track record should pilot it carefully.
Who should use it
Founders, small businesses and agencies with recurring work across Gmail, Slack, a CRM, Stripe or social networks who want AI coworkers that run on schedules and ask before acting; people who already pay for Claude or ChatGPT and want that plan to power background work; and technical teams that want an open-source, self-hostable version of a hosted AI-employee product.
Who should skip it
Enterprises that need per-agent isolation, scoped API keys and a long vendor track record; teams that want deterministic, step-by-step workflow automation rather than agents; and anyone who wants a free hosted tier or the full Cloud feature set without paying for Rerun Cloud.
Free trial
$0 for 5 days
Pro
$29/month
Billed monthly
Extra Coworkers add-on
$10/month per 5 coworkers
Billed monthly
Rerun credits add-on
$10/month per 1,000 credits
Billed monthly
Enterprise
Custom
Open source (self-hosted)
Free
Free tier limits: Rerun Cloud has a 5-day trial with up to 3 coworkers and 350 credits; a card is required and $0 is charged until day 5. There is no ongoing free hosted plan; the self-hosted open-source build is free.
Note: Rerun charges for capacity (coworkers), not seats or runs. The pricing page shows yearly billing at 20% off, and prices exclude tax. Per the billing docs, credits pay for Rerun-hosted models and paid data (1 dollar = 100 credits); runs on your own Claude, ChatGPT or OpenCode plan or API keys cost nothing on Rerun's side. LinkedIn Pro and WhatsApp Pro are $9 a month per account or number, with the first free for 5 days. Subscribing or adding an add-on during the trial ends the trial and bills the plan immediately. API access: the API and billing docs describe it as paid-plan only, while a November 3, 2026 changelog entry says the API now works during the trial.
Modalities
Text, Images, Voice input
API pricing
A single MCP endpoint (JSON-RPC 2.0, 53 tools) plus webhook trigger URLs, included with the plan; no REST API and no separate API pricing
Available models
Agency running client automations in one workspace
Boxes group coworkers by client at no cost and seats are unlimited, so an agency can run content, reporting and outreach coworkers for several clients and let teammates answer approvals, using a separate workspace when a client needs a hard boundary.
Founder-led sales prospecting
One coworker finds prospects (through paid data on Rerun Cloud or your own data provider), the next drafts a first email for each, and a third files the result, with an approval before anything is sent.
Back-office routines on a schedule
Cron tasks such as a weekday invoice chase or a Friday report run on the always-on machine while your laptop is shut, and a Stripe webhook trigger can handle failed payments as they arrive.
Self-hosted AI coworkers for a privacy-conscious team
The AGPL-3.0 build runs on your own Docker host with your own Supabase, keeps credentials in the engine's data directory, and can use local models through Ollama or vLLM.
Rerun vs. Lindy
Lindy is a closed-source, Slack-native AI teammate priced per user from $29.99 a month with team credits, connecting to 1,000+ apps and MCP servers. Rerun is open source and self-hostable, charges $29 a month per pack of five coworkers with unlimited seats, gives each workspace its own cloud machine with a browser, and can run on your own Claude or ChatGPT plan.
Rerun vs. Marblism
Marblism sells a fixed set of seven pre-built AI employees (SEO content, email, social, leads, websites, calls and contracts) in one closed-source subscription. Rerun lets you assemble any number of coworkers from templates or a plain description, choose the model for each, and self-host the whole thing.
Rerun vs. Grok Bot
Grok Bot also runs named bots on a persistent cloud computer that sign into your apps and ask for approval, but it comes with The X Subscription or eligible Cursor and SuperGrok plans and is not open source. Rerun is model-agnostic, sold per coworker pack, and available under AGPL-3.0.
Rerun vs. Claude Cowork
Claude Cowork is built into the Claude app, priced per person through Claude plans, and has one assistant carry out tasks on Claude models. Rerun runs a team of coworkers on a shared, always-on cloud machine with schedules, webhooks and approvals for the whole workspace, and it can use a Claude plan as one of its model sources.
Rerun vs. OpenClaw
OpenClaw is a free, open-source personal assistant you run yourself and talk to through messaging apps such as Telegram, WhatsApp and Slack, aimed at technical users. Rerun is a team product with a visual board, templates, approvals and a hosted cloud option, and its self-hosted build needs Docker and Supabase.
Rerun vs. Paperclip
Paperclip is an MIT-licensed control plane for running coding agents such as Claude Code, Codex and Cursor as a company with budgets and approval gates. Rerun builds and hosts its own no-code coworkers for business tasks rather than orchestrating external coding agents.
What is Rerun?
Rerun (rerun.build, by Reunit SA) is a platform for AI coworkers: agents with their own prompt, memory, skills, files and tools that work in your apps on a schedule, on a webhook or when you message them. Each workspace runs on a private cloud machine, coworkers can use a browser on their own screen, and they pause for your approval before consequential actions. A built-in guide called the Foreman can plan and build a team for you.
Is Rerun related to Rerun.io?
No. Rerun at rerun.build is the AI coworker platform from Reunit SA. Rerun.io is a separate company and product: an open-source SDK and data platform for logging, querying and visualizing multimodal data in robotics and physical AI.
How much does Rerun cost?
Rerun Cloud has one plan: $29 a month for 5 coworkers, unlimited seats, a private machine and 500 credits a month, with yearly billing shown at 20% off. Extra coworkers cost $10 a month per 5, and 1,000 more credits cost $10 a month. A 5-day trial with 3 coworkers and 350 credits requires a card but charges $0 until day 5. Runs on your own Claude, ChatGPT or OpenCode plan or API keys use no Rerun credits. Enterprise pricing is custom, and the self-hosted open-source build is free.
Can Rerun use my Claude or ChatGPT subscription?
Yes. You can sign in with a Claude Pro or Max plan or a ChatGPT Plus or Pro plan, and coworkers run on it; the engine ships the official Claude Code and Codex command-line tools for this, and those runs use no Rerun credits. A subscription is optional: every Rerun Cloud plan includes credits for Rerun's hosted models. You can also connect an OpenCode key, API keys for Anthropic, OpenAI, Google or OpenRouter, or an OpenAI-compatible endpoint such as Ollama, LM Studio or vLLM, and pick the model per coworker.
Is Rerun open source, and can I self-host it?
Yes. Since November 3, 2026, the app and engine are on GitHub (rerun-build/rerun) under AGPL-3.0. Self-hosting uses Docker Compose and needs Docker, Node.js 22, a Supabase project and, for the default models and Foreman onboarding, an OpenRouter key. The first workspace runs on the same machine; each additional workspace needs its own server. There are no coworker or seat limits on your install.
What is different between Rerun Cloud and the self-hosted version?
The coworkers, chat, memory, skills, schedules, MCP and Composio support are the same code. The open-source build leaves out the template library and creator program, included model credits, the LinkedIn Pro and WhatsApp Pro apps, paid data, Rerun's own email notifications and invitations, the weekly recap and the iOS app, and it gets community support only. Rerun Cloud also provisions each workspace's machine for you.
How does Rerun stop a coworker from doing something I did not want?
A coworker can stop its turn to ask a question, request approval for a list of actions (approved or declined one by one), ask you to connect an app, or collect an API key through a masked form. Unattended scheduled and webhook runs pause the same way and resume when anyone on the team answers. You can also switch off tool families such as shell or browser per coworker, and turn off its ability to rewrite its own prompt or install tools.
Where do Rerun coworkers run, and where is my data?
On Rerun Cloud, each workspace gets its own always-on machine in the cloud, in Europe according to Rerun's FAQ, separate from other customers. Coworkers' files, memory, databases and app credentials stay on that machine, and Rerun says it does not train models on your content. All coworkers in one workspace share that machine, so use separate workspaces when you need a hard boundary. On a self-hosted install, everything runs on your own servers and Supabase project.
Rerun vs Lindy: what is the difference?
Both give you hosted AI agents that work across your apps with triggers and approvals. Lindy is closed source, priced per user with credit metering, and centered on a Slack-native assistant. Rerun is open source and self-hostable, prices by packs of coworkers with unlimited seats, gives each workspace its own cloud machine with a browser, and can run on your own Claude or ChatGPT plan or API keys.

Lindy
Individual professionals and small teams who want an AI assistant that works across their actual email, calendar, and CRM without building explicit automation flows
Paid
Marblism
Solo operators and very small teams who need to cover multiple business roles (content, admin, social, sales, calls, legal review) and want a bundled AI solution at a lower cost than separate tools or part-time hires
PaidSpaceXSI
Individuals and teams on The X Subscription (Premium or higher) or existing Cursor and SuperGrok plans who want delegated, always-on agents that act in their real web apps and ask before anything consequential goes out
Paid
Anthropic
Professionals with recurring document, spreadsheet, or research workflows who already pay for Claude and want them done autonomously
PaidRerun is built around three ideas: a workspace, a Box and a coworker. A workspace is your account, and it runs on one private computer in the cloud (in Europe, per Rerun's FAQ) that every coworker in it shares, along with a shared folder, a shared database and the credentials of connected apps. Boxes are free groupings on the isometric board, by team, client or project. A coworker is an AI worker with a name, a markdown system prompt Rerun calls its soul, a model, its own memory, skills, files and SQLite database, and a set of tool families you can switch off (files, shell, web, browser, scheduling, webhooks, delegation and so on). You start by telling the Foreman, Rerun's built-in guide, what you need: it interviews you, proposes a team and builds it, or you install a ready-made template (executive assistant, SDR, social media manager, collections and more) or describe a single coworker and let it write its own prompt. Coworkers run on a cron, on a public webhook URL, when another coworker calls them by @handle, or when you chat with them, and they can split work into up to four parallel sub-coworkers. Since October 2026 each coworker also works on a screen of its own, with a browser, files and a terminal, so it can click through sites that have no connector while you watch or take over. Apps come from Rerun's catalog or from any MCP server you register: Rerun's site advertises 3,000+ apps and MCPs, while the open-source catalog seed lists 216 services, with Composio adding 2,000+ more apps by OAuth. Human-in-the-loop is built in: a coworker can pause mid-run, including at 3am on a scheduled run, to ask a question, request approval for a list of actions, ask you to connect an app, or collect an API key through a masked form whose value never reaches the model. The model layer is unusually open: Rerun's hosted Fast, Balanced and Smart tiers paid from included credits, a Claude Pro or Max or ChatGPT Plus or Pro plan (the engine ships the Claude Code and Codex CLIs to use them), an OpenCode key, API keys for Anthropic, OpenAI, Google or OpenRouter, or any OpenAI-compatible endpoint such as Ollama or vLLM. Rerun Cloud costs $29 a month for five coworkers with unlimited seats and 500 credits, after a 5-day card-required trial, with extra coworkers sold in packs of five. On November 3, 2026, the two-founder team open-sourced the app and engine under AGPL-3.0 with a Docker Compose install that needs a Supabase project. The tradeoffs: the self-hosted build drops the template library, included credits, LinkedIn and WhatsApp Pro apps, paid data, email notifications and the iOS app, and Google and Microsoft apps there need your own Composio setup; coworkers in a workspace share one machine and database, so Boxes are not a security boundary; coworkers may rewrite their own prompt and install tools unless you turn that off; and the product is young.
Coworkers can run on the models of an OpenCode Go plan, plus OpenCode Console models paid from OpenCode credits, connected with a single key. Rerun shows which models are included and tracks the plan's 5-hour, weekly and monthly limits.
Rerun published its app and the engine coworkers run on under AGPL-3.0 on GitHub, with a guide to host it with Docker Compose. Rerun Cloud keeps running for existing workspaces.
Two paid apps let coworkers work from a connected LinkedIn account (search, invitations, messages, posts) or reply to customers from a WhatsApp number, within each platform's daily limits. The first account or number is free for 5 days, then each costs $9 a month.
Each coworker got a screen with a browser, files and a terminal, so it can work on sites that have no connector. You can watch it, take over, and get it past sign-ins or captchas yourself.
Coworkers can look up and pay for leads, work emails, phone numbers, search and SEO data and social profiles with Rerun credits, with no separate data subscription. A later update (October 28) dropped the per-purchase approval and added a workspace switch to stop spending.
Are you the founder? Claim this listing →