Xero MCP Server

An MCP server for the Xero Accounting API (2.0). It lets an AI agent read one Xero organisation's contacts, invoices and bills, payments, chart of accounts and financial reports (profit and loss, balance sheet, aged receivables and payables) and, on explicit request, create contacts, invoices and payments or change an invoice's status.

• Transport: MCP Streamable HTTP at POST /mcp, health probe at GET /health

• Stateless and multi-tenant: the access token and organisation id arrive on every request as HTTP headers and are never stored; every tool is listed from the first tools/list

• Python 3.12, uv, FastMCP, httpx

• License: Apache-2.0, Copyright 2026 MSPbots.ai

Authentication headers

Where to get them:

• Create an app in the Xero developer portal (developer.xero.com, My Apps -> New app) and authorise it for the organisation with the OAuth 2.0 authorisation-code flow. Request the scopes the tools need: accounting.contacts, accounting.transactions, accounting.settings and accounting.reports.read (the .read variants for read-only use).

• The token response holds a short-lived access token; whoever stores the connection must refresh it with the refresh token. This server never refreshes or stores tokens.

• The tenant id comes from GET https://api.xero.com/connections with that token; each organisation the user authorised has its own tenantId.

How credentials are sent to Xero: every call goes to https://api.xero.com/api.xro/2.0/... with Authorization: Bearer <token> and xero-tenant-id: <tenant id>. Writes also carry a fresh Idempotency-Key header, so a retried write cannot create a second record. Values that look like unresolved templates (${...}) are treated as missing.

A request to /mcp without both headers gets HTTP 401 naming the missing one.

Related MCP server: Xero MCP Server

Tools

All tools return a JSON string. Ids are UUIDs. List tools return 100 records per page with {count, page, has_more, next_page}. Dates are YYYY-MM-DD; Xero's legacy /Date(...)/ values in responses are converted to ISO 8601.

Reports come back flattened as {name, titles, date, columns, sections: [{title, rows: [{label, values}], total}]} instead of Xero's nested rows and cells.

Errors come back as a JSON envelope instead of an exception:

{"error": {"code": "invalid_argument", "message": "Xero returned HTTP 400 for Invoices: A validation exception occurred - Account code '999' is not a valid code", "retryable": false}}

code is one of not_configured, unauthorized, not_found, invalid_argument, rate_limited, upstream_error; Xero's validation messages are included. Xero allows 60 calls per minute and 5,000 per day per organisation; 429, 5xx and network errors are retried with backoff (honouring Retry-After), and writes are safe to retry because of the idempotency key.

Run locally

uv sync
uv run python -m xero_mcp          # listens on 0.0.0.0:8080

or with Docker:

docker build --platform linux/amd64 -t xero-mcp:dev .
docker run --rm -p 8080:8080 xero-mcp:dev

Configuration (environment variables, all optional): MCP_HTTP_HOST (default 0.0.0.0), MCP_HTTP_PORT (default 8080), LOG_LEVEL (default INFO). Credentials are never read from the environment.

Testing example

# list tools
curl -s http://localhost:8080/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "X-Xero-Access-Token: your-access-token" \
  -H "X-Xero-Tenant-Id: your-tenant-id" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

# unpaid sales invoices
curl -s http://localhost:8080/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "X-Xero-Access-Token: your-access-token" \
  -H "X-Xero-Tenant-Id: your-tenant-id" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"xero_invoices_list","arguments":{"Type":"ACCREC","Status":"AUTHORISED"}}}'

Unit tests (no network access needed):

uv run pytest

Delivery notes

• Tool count (20). Kept as is because the tool names are an existing contract.

• Aged reports. Xero's aged receivables / payables reports require a contact (contactId), and the API has no all-contacts version. With contactId the tools return Xero's report; without it they return a per-contact ageing summary (current, 1-30, 31-60, 61-90, over 90 days past due) built from open authorised invoices dated on or before date, using today's amount due.

• Searching. xero_contacts_search uses Xero's searchTerm parameter, so names with quotes are found and cannot break the filter; other filters escape their values.

• Write tools. xero_contacts_create, xero_invoices_create, xero_payments_create and xero_invoices_update_status change the books; their descriptions tell the agent to use them only on explicit request. Status changes are annotated destructive because voiding cannot be undone. Per-record validation errors that Xero returns with HTTP 200 are reported as errors.

• No interactive prompts. Older versions asked the user for filters before unfiltered lists, which a stateless gateway cannot relay; lists now page instead.

License

Apache License 2.0 - see LICENSE and NOTICE.