Link Claude, ChatGPT or any MCP-compatible AI tool to your Avaza account in a couple of minutes. Then ask it, in plain language, to plan projects, log time, book your team, draft invoices and report on the numbers, all within your own Avaza permissions.
https://mcp.avaza.com/mcpOn claude.ai and the Claude desktop app. The Free plan can add one custom connector; Pro, Max, Team and Enterprise can add more.
Avaza, paste the server URL https://mcp.avaza.com/mcp and click Add. You can leave the advanced settings empty.On Team and Enterprise plans an Owner adds the connector once, under Organization settings › Connectors. Each member then clicks Connect under Customize › Connectors and signs in with their own Avaza login.
Claude's guide to custom connectorsOn ChatGPT for the web, as a custom app in developer mode. In Business, Enterprise and Edu workspaces an admin turns developer mode on first.
Avaza, paste https://mcp.avaza.com/mcp as the MCP server URL and choose OAuth for authentication.ChatGPT asks you to confirm before it changes anything in Avaza. Custom apps work in ChatGPT on the web, not in the mobile apps.
OpenAI's guide to developer mode and MCP appsAnthropic's coding agent, in your terminal or IDE.
--scope user makes it available in every project./mcp, choose avaza and sign in to Avaza in your browser.claude mcp add --transport http --scope user avaza https://mcp.avaza.com/mcpCursor's agent, on the desktop app.
~/.cursor/mcp.json to use it in every project, or to .cursor/mcp.json in one project.{
"mcpServers": {
"avaza": {
"url": "https://mcp.avaza.com/mcp"
}
}
}GitHub Copilot's agent mode in Visual Studio Code.
.vscode/mcp.json in your workspace, or run MCP: Open User Configuration to add it for every workspace.{
"servers": {
"avaza": {
"type": "http",
"url": "https://mcp.avaza.com/mcp"
}
}
}Using another tool? Add a remote (HTTP) MCP server with the URL above and sign in to Avaza when asked: sign-in is standard OAuth, so there's no API key to copy. AI tools update their menus often; if a step doesn't match what you see, the tool's own guide has the latest wording.
Your assistant chains Avaza's tools together: it looks things up, makes the changes and reports back. Put routine checks on a schedule and they run on their own.
Claude
ChatGPTClaude
Set up a website redesign for Bluegum Architects with Discovery, Design, Build and Launch phases, due 30 November. Book Priya and Leo at 4 hours a day.
Claude · Done in Avaza
create_projectWebsite Redesign for Bluegum Architectscreate_sectionDiscovery, Design, Build and Launchcreate_task14 tasks with start and due datesadd_schedule_bookingPriya and Leo, 4 hrs/day to 30 Nov
Website Redesign · Bluegum ArchitectsDue 30 NovProject ready: 4 phases, 14 tasks and two people booked through November.
ChatGPT
Log 6 hours a day on the Fernleaf Media campaign Monday to Wednesday and 4 hours on Harbour Legal on Thursday, then submit my timesheet.
ChatGPT · Done in Avaza
list_projectsMatched Fernleaf Media and Harbour Legalcreate_timesheet4 entries, 22 hourssubmit_timesheetsWeek of 14 Sep sent for approvalTimesheet · Week of 14 SepPending approval22 hours logged and submitted for approval.
Claude
Invoice Fernleaf Media for all unbilled time and expenses on the Marketing Campaign for September.
Claude · Done in Avaza
get_uninvoiced_summary38.5 hours and 2 expenses unbilledcreate_invoiceDraft invoice for Fernleaf Mediaadd_uninvoiced_to_invoiceTime and expenses added as line itemsInvoice · Fernleaf MediaDraftDraft invoice ready for your review, nothing sent until you say so.
ClaudeSummarise last week's hours by project, flag any project past 80% of its budget, and list anyone who hasn't submitted a timesheet.
Claude · Done in Avaza
timesheet_summaryLast week's hours by projectlist_projectsBudget used on every active projectlist_timesheetsUnsubmitted time foundMonday briefingScheduled runBriefing ready: 312 hours logged, 2 projects past 80% of budget, 3 timesheets still to submit.
Create projects and tasks, log time and expenses, book people, raise invoices and approve timesheets by describing what you need, in the AI tool your team already uses.
Claude can run scheduled tasks hourly, daily or weekly, with your connectors. Write an Avaza routine once and your briefings, timesheet checks and month-end billing run on their own.
Your AI tool can combine Avaza with its other connectors, such as email, calendar and documents. Zapier MCP reaches thousands more apps, and webhooks tell other systems the moment something changes in Avaza.
The MCP server covers the whole project-to-payment cycle. Your assistant picks the right tools for each request, so you never need to know their names.
Create projects, sections and tasks, add team members, update statuses and due dates, and comment on tasks.
Log time, start and stop timers, submit timesheets and approve or reject your team's entries.
Record and submit expenses, look up categories and merchants, and summarise spending.
Turn unbilled time and expenses into draft invoices, add line items and record customer payments.
Create, revise and review quotes with line items, taxes and discounts.
Enter supplier bills with line items and record bill payments.
Book people onto projects, add leave and see who is booked when.
Create and update companies and contacts, and look up anyone on your books.
Summarise hours and expenses by project, person or category, and see what's ready to bill.
Click a prompt to copy it, then paste it into your AI tool. Swap in your own project, client and people names.
Deletions always come back as a preview first and only go ahead after a separate confirmation within five minutes. Your AI tool may run that step itself, so keep it set to ask before it acts.
The MCP server uses the same logins, roles and permissions as Avaza itself, with extra admin controls built for AI.
Each person connects their own account on Avaza's sign-in page. Passwords are never shared with the AI tool, and access can be revoked at any time.
The assistant sees and changes only what your Avaza role allows. If you can't open a project in Avaza, neither can your AI tool.
Admins set Read, Create, Update and Delete access for each module (role-based, admins only or blocked), tighten it for individual users and cap bulk deletions, under Settings › Developer Apps & API Access › API permissions.
Every deletion comes back as a preview first and only goes ahead once it's confirmed. Confirmations expire after five minutes.
MCP (Model Context Protocol) is an open standard that lets AI assistants use other apps' tools securely. The Avaza MCP Server gives AI tools such as Claude and ChatGPT 121 tools for working with your Avaza account, so they can look things up and take action when you ask.
Any paid Avaza plan. Each person connects with their own Avaza login and works within their own role and permissions.
Claude and ChatGPT, plus coding tools such as Claude Code, Cursor and VS Code, and any other tool that supports remote MCP servers with OAuth sign-in. Your AI tool's own plan decides whether custom connectors are available; Claude's Free plan, for example, allows one.
No. It works as the person who connected it, sees only what their Avaza role allows, and fetches only what it needs for each request. Admins can restrict it further with API permissions.
Yes. Under Settings › Developer Apps & API Access › API permissions, admins choose Read, Create, Update and Delete access for each module: role-based, admins only or blocked. They can make it stricter for individual users and cap how many deletions each person can make per hour. The same rules apply to personal API tokens and connected apps.
Deletions take two steps: the assistant first gets a preview of what would be removed, and nothing happens until it's confirmed. Confirmations expire after five minutes, some records are protected (such as estimates already converted to invoices), and admins can block deletions altogether.
Only the Avaza data your assistant fetches while working on your request. How the AI provider stores and uses your conversations is covered by its own terms and your plan with them, so check your AI tool's data settings.
Sign-ins expire from time to time for security. Reconnect Avaza in your AI tool's connector settings and sign in again. If it still fails, check that the URL is exactly https://mcp.avaza.com/mcp, with no trailing slash, and that your browser allows the sign-in pop-up.
Check that the name matches what's in Avaza and that your Avaza user can see it. The assistant has exactly the same visibility as you.
Remove Avaza in your AI tool's connector settings (in Claude, Customize › Connectors). Admins can also block API access to any module in API permissions.
Connect in a couple of minutes, or start with a free Avaza account and explore first.