CorebillTell it what you need.
It handles the paperwork.
Search customers, draft quotes, send invoices, record payments and build reports — just by asking. It searches your catalog, uses your real prices and taxes, and confirms before it creates or sends anything.
24 real tools, not a chat toy.
Every action below runs against your own data — the same customers, quotes and invoices you see in the app.
Customers
Finds, creates and updates customers, with slugs generated automatically.
search_customers · create_customer · update_customerQuotes
Drafts quotes with your catalog, creates new versions, sends and converts them to invoices.
create_quote · send_quote · convert_quote_to_invoiceInvoices
Creates full or partial invoices, sends them and records payments as they come in.
create_invoice · send_invoice · record_paymentServices & catalog
Looks up an item by name before asking you for a price — it uses what's already in your catalog.
search_items · list_items · create_itemPayments & Stripe
Lists subscriptions and transactions, and can pause, resume or cancel a Stripe subscription.
list_stripe_subscriptions · cancel_stripe_subscriptionAnalytics & reports
Answers numeric questions on the spot, or builds a chart panel in your Reports dashboards.
get_revenue_by_period · create_report_panel · run_reportIt answers in the language you write in.
No settings to change: the agent detects the language of each message and replies in kind, whether your team writes in English or Español.
Fast, but never reckless.
The agent is built to act like a careful teammate: it asks, it stays in its lane, and it leaves a trail.
Confirms before it acts
Before it creates, sends or records anything, it shows you the details and waits for your go-ahead.
Respects your team's roles
A viewer can search and ask questions, but write actions — create, update, send, record a payment — are blocked for that role.
Scoped to your company
Every tool call runs against one company at a time. It never reads or touches another workspace's data.
Nothing happens off the record
Every message and every tool call is saved to the conversation, so your team can review exactly what it did.
One agent, wherever you're already looking.
Keep it open next to whatever record you're on, or give it a full page when you need a longer back-and-forth.
Side panel, anywhere
Press ⌘ J (or Ctrl J) from any screen to open it without losing your place.
A full page when you need it
Every conversation is saved with its own history, so a longer session — planning a batch of invoices, digging through analytics — is never lost.
Whatever you'd tell an assistant, write it here.
No syntax to learn — every audience gets the same plain-English agent.
Service businesses
- “Remind everyone with an invoice over 15 days late”
- “What did we bill last month by service?”
- “Create a customer for Blue Harbor and quote the onboarding”
Digital agencies
- “Quote the Q4 retainer for Northwind Labs, 20 hours a month”
- “Invoice the 30% deposit for the Kappa Retail project”
- “Show me every open quote waiting on a client”
Freelancers
- “Send Laura a quote for the logo package”
- “Has Acme Studio paid the last invoice?”
- “Which invoices are due this week?”
Startups
- “Chart monthly revenue for this year, split by plan”
- “How many customers churned last quarter?”
- “Pause the subscription for Orbit Media”
Developers
- “List the invoices created by the API this week”
- “What's our AI message usage on the current plan?”
- “Find the customer linked to stripe_customer cus_9f2c”
The same agent, over your own API.
POST /v1/ai/chat streams the exact same tool calls you see in the app — text, tool calls and results — as Server-Sent Events, so your CLI or product can render them live.
Questions about the agent
Corebill runs on Claude, Anthropic's model family. The free plan uses the fast Haiku model; paid plans can pick a more capable model for a given conversation.
Every plan includes a monthly allowance of AI messages, shown right in the agent panel. If you need more, you can upgrade your plan at any time.
No. Before it creates, sends or records anything, it shows you the details and waits for your confirmation — it never fires off an action with side effects on its own.
Yes — give them the viewer role. Viewers can still search and ask questions, but every action that creates, updates or sends something is blocked for that role, in the agent and everywhere else.
Whatever language you write in — it replies in the same one. Your organization can also set a default interface language separately, per workspace, in Settings.
Try it on your own quotes and invoices.
Create your workspace, add a customer, and ask the agent to draft the first quote with you.