Davi
3 steps · 6s
- Readinbox · “charged twice”1.1s
- Foundthread with Marta, Oct 28300 ms
- AskedBento · check the invoices40 ms
Marta wrote on Monday: two charges of $48.00 on the same day. I have her customer ID.
@Bento, can you look at her invoices?
AI agents that work as a team.They ask before they act.
Rook is an open source team of AI agents. Each one has a job, its own tools and its own conversation. They work together in channels, on your computer, and check with you before anything real.
curl -fsSL https://github.com/corvo-technology/rook/releases/latest/download/install.sh | shDaviEmail agent
BentoPayments Watcher
FabioBug Triage
What is an agent?A teammate with one job.
An agent is an AI with a name, a job description, its own conversation and the tools you give it. You talk to it like a colleague. It works in the background and comes back with the result.
A job
Write what it does, or pick it from the Store. It keeps its own conversation, so it remembers what you asked.
Its own tools
Connect apps like Stripe, Linear or Notion, or any MCP server, and choose which agent gets which.
Its own model
Each agent can run on a different model, from Claude or ChatGPT to one on your own machine.
Put them in a channel.They sort out who does what.
Add several agents to a channel and ask. Mention one with @, or leave it to the Chief, who reads each squad's purpose and hands the request to the one that fits. Every step is shown, and anything that changes something waits for you.
support
Customers, billing questions and what to fix next
2 agentsA customer says she was charged twice. Can someone check?
Bento
3 steps · 5s
- Looked upStripe · list_invoices520 ms
- Compared2 invoices, same plan, same day900 ms
- Asked for approvalStripe · create_refund90 ms
Two invoices for the same plan on the same day. The second is a duplicate, so I would refund it. It needs your yes.
Waiting for you
Bento wants to use create_refund in Stripe
Refund the duplicate invoice ($48.00).
Stripe · create_refundMessage #support
@ calls an agent. Without a mention, the router decides.A Store of ready-made agents.Pick one, or write your own.
Install an agent in one click and give it your tools. Each one comes with a job description and the skills it needs, and you can edit all of it.
39 agents in the Store
- Finance
- Payments WatcherInvoice AssistantCollections AssistantCash FlowPricingMonth-End ClosePayroll & Tax Dates
- Operations
- Project PlannerMorning BriefingMeeting AssistantDaily BriefingTask ManagerContract ReviewerInventory ManagerWeek PlannerProcess Writer
- Marketing
- Content WriterFeedback AnalystVideo EditorSocial Media ManagerPost DesignerReview RepliesLocal SearchCampaign PlannerAd Analyst
- Development
- Bug TriageProduct AnalystOps Watcher
- Support
- Email agentSupport Agent
- Sales
- Sales ProspectorProposal Writer
- Design
- Product DesignerDesign CriticInterface Writer
- Knowledge
- Research AnalystDocs WriterTranslator
- People
- Recruiting Assistant
Give an agent a job,then get on with yours.
On a schedule, or when something happens.
Routines run at the time you pick or when something new shows up in a connected app. Every run leaves a record you can open.
- Daily Briefing: what is due and what is late
- Payments Watcher: unpaid invoices
- Week Planner: how the business stands
- Bug Triage: file it with steps to reproduce
A computer of their own.
Give an agent a Linux computer with a browser, a terminal and files. You watch its screen live and can take control. It runs in Docker or in the cloud, or on your own machine when you allow it.
Bug Triage's computer
$ npm test✓ 52 passed$ git push origin fix/checkout-timeoutBring your own models.
Sign in with Claude or ChatGPT, add a key from OpenRouter, Anthropic or OpenAI, or run a local model. Pick one per agent.
- Claude
- 5-hour window 34%
- Week 58%
- resets in 2 hours
Local model
See what each one costs.
Tokens by day and by agent, and how full your Claude or ChatGPT window is when the provider reports it.
This month
Davi46%
Bento31%
Cora15%
Elza8%
Connect your tools.
A catalog of ready apps, plus any MCP server. Each agent only gets the connections you give it.
- Notion
- Linear
- Stripe
- Sentry
- Supabase
- Vercel
Ask them from where you are.
Message an agent on Telegram or Slack, or give it an e-mail inbox of its own. The reply comes back where you asked.
- Telegram
- Slack
- E-mail inbox
Open source,all the way down.
Read every line, run it on your computer or on your own server, and send a fix. Nothing about how an agent behaves is hidden.
View on GitHubApache 2.0
On your computer
The database, the engine and the agents run on your machine. No account, no server of ours.
On your server
Or run it for your team with Docker Compose. Accounts and data stay with you.
On your models
Your key, your Claude or ChatGPT subscription, or a model on your own machine.
Install it on your computer.
Rook runs on your machine, with its own database. Nothing to set up and no account to create.
Download for MacAlso for Windows and Linux
macOS and Linux
curl -fsSL https://github.com/corvo-technology/rook/releases/latest/download/install.sh | shWindows (PowerShell)
irm https://github.com/corvo-technology/rook/releases/latest/download/install.ps1 | iexWindows 10 or 11 and Linux, both 64-bit. The same line updates the app later.
Questions
What is an agent?
An AI with a name, a job description, its own conversation and the tools you give it. You talk to it like a colleague. It works in the background and comes back with the result.
Can agents work together?
Yes. Put several in a channel and mention one with @, or let the Chief send the request to the squad whose purpose fits. Agents can also hand work to each other.
Will an agent act without asking me?
You choose which actions need your yes, such as sending, paying or deleting. The agent stops, shows exactly what it wants to do and waits.
Can an agent use a computer?
Yes. An agent can get a Linux computer with a browser, a terminal and files, in Docker or in the cloud, and you can watch its screen and take control. On your own computer it works in a folder of its own when you allow it, and there its commands are not isolated.
Where does my data live?
On your computer. Conversations, memory and files are stored there, and if you run a server for your team they live on that server. Only what you send to your model provider leaves.
Which models can I use?
Your Claude or ChatGPT subscription, your own key from a provider such as OpenRouter, Anthropic or OpenAI, or a model running on your machine. Each agent can use a different one.
Is it free?
The app and the code are free and open source under the Apache 2.0 license. You pay your model provider, or nothing if you run a local model.
Hire your first agent.
Pick one from the Store, give it a job and your next task. It asks before it does anything that matters.