LaunchedEditorial Listing

Rerun

Reunit SA · Rerun: Open-Source AI Coworkers on Schedules, With Their Own Cloud Computer

Open Rerun

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.

PricingFreemium
Setupeasy
Runs onWeb · API · Self-hosted
APIYes
Open sourceYes
DocsYes
CategoryProductivity
AI EmployeesNo-CodeAgent BuilderWorkflow AutomationScheduled TasksWebhooksHuman-in-the-LoopComputer UseCloud ComputerAlways-OnMulti-AgentPersistent MemoryMCPBYOKModel-AgnosticClaudeChatGPTOpenRouterOllamaOpen SourceSelf-HostedDockerSmall BusinessAgencies

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

Capabilities

  • The Foreman, a built-in guide that interviews you about the work, proposes a team of coworkers and builds it, and can answer coworkers by rules you set while you are away
  • Coworkers with their own system prompt (soul), model, memory, skills, files and private SQLite database, plus a folder and database shared across the workspace
  • Scheduled tasks on a 5-field cron with an IANA timezone, one-off dated runs, a per-task model, and a Run now button for testing
  • Webhook triggers: a public URL per trigger that starts a run with the caller's payload attached
  • Human-in-the-loop pauses for questions, per-action approvals, app connection and masked credential forms, which also work on unattended scheduled and webhook runs, with notifications in the app and by email
  • A private cloud machine per workspace that stays on, where each coworker gets a screen with a browser, files and a terminal that you can watch and take over
  • Apps installed per coworker from a catalog (216 services in the open-source seed, 2,000+ more via Composio), with several accounts of the same app per coworker, plus any HTTP, SSE or local-command MCP server
  • Bring your own model: Claude Pro or Max and ChatGPT Plus or Pro plans, OpenCode Go and Console, API keys for Anthropic, OpenAI, Google and OpenRouter, or a custom OpenAI-compatible endpoint (Ollama, LM Studio, vLLM and others), alongside Rerun's hosted Fast, Balanced and Smart tiers
  • Coworkers call each other by @handle and can spawn up to four throwaway sub-coworkers in parallel
  • Per-coworker capability switches (filesystem, shell behind a denylist, web, browser, memory, scheduling, triggers, delegation) and self-improvement flags for prompt edits, tool installs and automatic memory
  • Template library with ready-made coworkers, share links, and a marketplace where creators sell templates from $5 to $2,000 (Rerun keeps 5% + $0.30 per sale)
  • Paid data on Rerun Cloud: leads, work emails, phone numbers, SEO metrics and social profiles bought with Rerun credits, with sources shown per field
  • Monitoring and Logs tabs that record every run's steps, tool calls, tokens and cost
  • MCP API endpoint with 53 tools to create and drive coworkers, skills, schedules, triggers, apps, databases and runs from Claude Code, Codex, Cursor or scripts
  • Open source under AGPL-3.0 with a Docker Compose install; extra workspaces attach their own engine server

Limitations

  • Rerun Cloud has no free plan: the 5-day trial needs a card and runs only 3 coworkers at once on 350 credits
  • The open-source build leaves out the template library, included model credits, LinkedIn Pro and WhatsApp Pro, paid data, notification and invitation emails, the weekly recap and the iOS app, and gets community support only
  • Self-hosting needs Docker with about 6 GB of memory for the image build, Node.js 22 and a Supabase project, and every workspace beyond the first needs a server of its own
  • On a self-hosted install, Google and Microsoft apps, Instagram, LinkedIn, Reddit and YouTube connect only after you set up your own Composio project and an auth config per service
  • All coworkers in a workspace share one machine, folder and database, so Boxes do not isolate them, and any builder in the workspace can ask a coworker to act on accounts connected to it
  • Coworkers may rewrite their own system prompt and install packages and MCP servers by default; both must be switched off for a coworker that should stay exactly as written
  • Webhook trigger URLs carry no signature or secret header: anyone holding the URL can start a run
  • No REST API: everything goes through a single MCP endpoint, approval cards can only be answered in the app, and API keys are workspace-wide with no read-only scope
  • LinkedIn Pro and WhatsApp Pro cost $9 a month per account or number on top of the plan, after 5 free days for the first one
  • Local models need care: the docs warn that Ollama's default 4k context truncates a coworker's roughly 28k-token instructions and that small models are uneven at tool calls
  • A young product from a two-founder team, open-sourced on November 3, 2026

Use cases

  • Having an executive-assistant coworker sort the inbox, clear the calendar and send a morning brief every weekday at 7am
  • Chaining coworkers so one finds prospects, the next writes each a first email and a third files the results in the CRM, with approval before anything is sent
  • Chasing late invoices on a schedule and handling Stripe events through a webhook trigger
  • Drafting a week of social posts per network that publish only after you approve them
  • Running client automations for several clients from one agency workspace, grouped into Boxes
  • Building and versioning coworkers from Claude Code or Cursor through Rerun's MCP endpoint

Our take

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.

Strengths

  • Runs on the Claude, ChatGPT or OpenCode plan you already pay for, your own API keys or a local model, with no Rerun charge for those runs
  • Open source under AGPL-3.0 and self-hostable with Docker Compose, with no coworker or seat limits on your own install
  • Flat pricing per pack of five coworkers with unlimited seats and no per-run fees
  • Approvals, questions and credential requests that pause even unattended scheduled runs, with secrets kept on the workspace machine and out of the model's context
  • Each workspace has its own always-on machine with a browser and screen you can watch and take over
  • The Foreman plans and builds a team from a plain-language interview, and ready-made templates cover common roles

Weaknesses

  • No free hosted plan, and the trial needs a card
  • Self-hosted installs lose the template library, paid data, LinkedIn and WhatsApp Pro and email notifications
  • Coworkers in one workspace share a machine and database, with no per-coworker isolation or scoped API keys
  • Unsigned webhook URLs and self-modifying coworkers by default call for some care
  • Young product from a very small team

Rerun pricing

Free trial

$0 for 5 days

  • Up to 3 coworkers at once
  • 350 credits
  • Unlimited seats
  • Card required; the plan starts on day 5 unless cancelled

Pro

$29/month

Billed monthly

  • 5 coworkers
  • Unlimited seats
  • Private cloud machine per workspace
  • 500 credits every month
  • Every app, template and unlimited runs
  • Smart model tier

Extra Coworkers add-on

$10/month per 5 coworkers

Billed monthly

  • Added to the same subscription
  • Removable at the next renewal

Rerun credits add-on

$10/month per 1,000 credits

Billed monthly

  • Credits for Rerun-hosted models and paid data
  • 1 dollar = 100 credits

Enterprise

Custom

  • Custom or self-hosted infrastructure
  • Dedicated support
  • SSO
  • SLA

Open source (self-hosted)

Free

  • AGPL-3.0
  • Unlimited coworkers and seats
  • You pay your model provider and host directly
  • Community support

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.

Technical specs

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

Rerun hosted tiers: Fast, Balanced and Smart (on Rerun credits, or on your OpenRouter key when self-hosted)Claude via a Claude Pro or Max plan or an Anthropic API keyGPT models via a ChatGPT Plus or Pro plan or an OpenAI API keyGemini via a Google API keyAny OpenRouter modelOpenCode Go and Console modelsOpenAI-compatible endpoints: Ollama, LM Studio, vLLM, LiteLLM, Groq, Together, Mistral, DeepSeek

Where Rerun excels

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. competitors

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.

Frequently asked questions

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.

Integrations & fit

GmailGoogle CalendarGoogle DriveGoogle SheetsSlackNotionHubSpotAirtableStripeShopifyGitHubLinearJiraFigmaWebflowLinkedInWhatsAppInstagramXYouTubeComposioMCPWebhooksClaudeChatGPTOpenCodeOpenRouterOllamaClaude CodeCodexCursor
Good fit forSolo / individual, Startup / small team
Pricing modelFreemium· Free tier available
See pricing on Rerun →

Alternatives to consider

About Rerun

Rerun 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.

Updates from Rerun

IntegrationOpenCode subscriptions as a model source

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.

LaunchRerun open-sourced under AGPL-3.0

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.

IntegrationLinkedIn Pro and WhatsApp Pro apps

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.

New FeatureA computer of their own

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.

New FeatureCoworkers can buy leads, SEO and social data

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 →