The Payhawk MCP handles multi-step finance work through playbooks and Payhawk AI agents.
Playbooks are multi-step procedures for finance jobs that take several actions in the right order for your AI Assistant to follow.
AI agents run longer jobs in the background, such as deep spend analysis, without blocking the chat.
Available built-in playbooks
The built-in Payhawk playbooks are readily available for you to trigger. You don't have to remember the steps - you can ask for the job, and your AI Assistant will follow the playbook, calling the right tools as it goes.
A playbook is offered based on the permissions its steps need, not your role name, so you only ever see the playbooks you can run.
Approval triage works your own approval queue.
Custom roles are treated on the permissions they carry, not the role they were cloned from.
The following table lists the built-in playbooks Payhawk currently supports.
Playbook | What it does |
|---|---|
Benchmark against peers | Whether your spend mix, expense process, and product adoption are normal for a company like yours, and which of your entities is lagging. |
Book a flight | Searches flights with your company travel policy applied and submits a policy-checked flight request into your account's normal approval workflow. Owned and handled natively by the Travel agent; nothing is purchased in chat. |
Book a group trip | Resolves who is travelling, adds or removes travellers on one trip, and syncs their details. |
Book a hotel | Searches stays with your company travel policy applied and submits a hotel request into your account's normal approval workflow. Owned and handled natively by the Travel agent; nothing is purchased in chat. |
Book a multi-city trip | Records every stop so each gets its own hotel search, and lets you park a stop and resume it. |
Book a trip for someone else | Plans and books a full trip on behalf of another employee, using their traveller profile and policy limits |
Check the next payment run is covered | Whether the money will be there to pay the bills falling due by a deadline, per currency, and if not how much to load and by when. Checks funding. Never pays anything. |
Fix inconsistent coding | Where the same supplier, employee, spend type, or category is coded inconsistently, what it needs to default to, and the setup change that keeps future coding right. |
Forecast budget overspend | Which budgets are trending to overspend by period end, and why. |
Forecast when to top up | Forecasts when your balance runs low at the current spend rate, per currency and entity, so you top up in time. Never moves money. |
Manage a trip | Views, changes or cancels an existing trip and its bookings. Owned and handled natively by the Travel agent, with changes going through your account's normal approval workflow. |
Measure the finance process | Finance process timing from workflow timestamps - time to submit, approve, and review per team, and where the close is bottlenecked. Team-level by default. |
Merge duplicate suppliers | Takes a vendor from several supplier records down to one - re-points what can move, archives the redundant records with a visible trace, and reports what could not follow. |
Prepare a leaver handover | Everything a leaver's role empowers and owns - cards, spend, approvals, workflows, budgets, teams - ordered by what breaks first and phased around the last day. |
Profile a supplier's spend | The complete spend picture for one supplier - total and trend, open commitments, unpaid bills, and recurring charges, including spend on old supplier records a manual total would miss. Read-only. |
Report cash position | Your current cash position per balance type, currency, and entity, with recent movement and the balance trend where history is available. |
Review card and spend-control fit | Finds where card limits and spend policies no longer match how people actually spend, then applies the fix at the level that fits, namely, a single card's limit, or the spend policy itself when a whole group has outgrown it. |
Review open expenses | Reviews a batch of expenses for coding, policy, and receipts, then approves or returns each one. |
Review unusual spend | Surfaces unusual spend and its likely drivers, then looks into the flagged items. |
Roll up spend across entities | Consolidates spend, balances, and activity across every entity you can access, grouped per entity. |
Run the month-end close | Gets the books ready to close - codes everything, chases receipts, clears your approvals, completes the review sign-off, and finishes with a shareable close summary. |
Subscription and Saas spend audit | Reviews your recurring and subscription spend and flags any price rises, then finishes the cleanup - confirms or dismisses the charges nobody has marked as subscriptions, and merges a supplier's duplicate records into one. |
Triage approvals | Everything awaiting your own approval across every entity you cover, in one complete ranked queue, acted on in context. Never shows the account-wide backlog. |
Write a spend report | Produces the spend report for any period you name, as one answer containing the narrative, the key figures and the charts. Merges monthly finance summary, spend dashboard and period comparison into one playbook. |
Custom playbooks
Besides the built-in Payhawk playbooks, you can create your own custom playbooks with the Payhawk MCP to meet your company's specific needs.
You build and update custom playbooks conversationally, by chatting with your AI Assistant. Once a playbook is ready, you run it just like a built-in one — by asking for it in the chat.
Running the playbooks
You don’t need a special command to run a playbook. Tell the AI Assistant the job you want to do in your own words, for example, "close the month" or "run the weekly expense review". Your AI Assistant will check the playbooks you can run, find the one that matches, and work through its steps with you.
For reliable results, follow the recommendations:
Ask what is available - Say "what playbooks do I have?" to see every playbook you can run, each with a short description.
A question does not start a playbook - If you ask something like "what is blocking the close?", the AI Assistant will answer the question. To run the full procedure, ask for the job itself ("close the month").
Invoking the Payhawk MCP Agents
The following table lists the AI Agents supported by the Payhawk MCP.
AI Agent | What it does |
|---|---|
Financial Controller | Works through month-end close, reviewing and reconciling expenses across the account. |
Payments | Answers questions about upcoming payment obligations. |
Procurement | Handles purchasing tasks, such as drafting purchase orders. |
Travel | Searches flights, hotels, and rail with your company travel policy applied. |
Working with travel bookings
When working with the Payhawk MCP, nothing is bought in the chat on any AI Assistant. The Travel AI Agent searches the options with your company policy applied and, then, submits a travel request into your company's approval workflow, the same way a regular request works. When the request is approved, Payhawk completes the booking. The flow is identical in the supported AI Assistants.
In ChatGPT, OpenAI's own safety checks can occasionally block a travel step with a message that the tool call was blocked. As this check comes from OpenAI, retry the request and state a clear business reason for the trip to resolve the issue.