Use the Xero MCP server — troubleshoot its self-refreshing OAuth2 token file, maintain the local best-of-breed server build, and pick up other operational notes for working with Xero MCP tools. Use when Xero MCP tools fail with authentication errors, before starting any Xero workflow, when adding or updating upstream xero-mcp-server PRs, or for general guidance on Xero MCP usage.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Covers usage of the Xero MCP server, including authentication (OAuth2 bearer tokens) and related operational notes.
For browser-driven automation of the Xero web UI (not via MCP), see the
separate xero-browser skill instead.
Before any Xero data task, read references/mcp-vs-browser.md.
It is the single source of truth for choosing MCP tools vs the browser
Account Transactions report (shared with the xero-browser skill — do
not duplicate its content here). Key rule: Xero MCP has no per-account
transaction endpoint, so a single account's full ledger comes from the
browser report, never from paginating list-manual-journals.
The server reads a self-refreshing OAuth2 token file on every tool call, so tokens never need refreshing by hand and the agent session never needs restarting for auth. Several agents (Claude Code, Pi, OpenCode) share the file safely: refreshes are serialised by a lock, so each expiry is refreshed exactly once. They also share Xero's rate limits — "Too many requests to Xero" is not an auth failure; avoid fanning out many parallel Xero calls and retry after a moment.
An "Authentication failed" / 401 from a Xero MCP tool is not a routine
expiry any more. Do not retry blindly, and do not switch to a browser /
block-explorer / Cryptio workaround "to keep moving": the MCP path is far
faster and more reliable. Instead, read
references/authentication.md — the
usual causes are a scope the token lacks (one tool fails, others work)
or a refresh token unused for 60 days (every tool fails), which needs the
user to run the browser OAuth flow.
The ONLY time a browser fallback is right is when the data genuinely
isn't available via MCP at all (e.g. a per-account transaction ledger —
see references/mcp-vs-browser.md).
Also read references/authentication.md
for first-time setup, after revoking and re-authorizing the Xero app, or
when changing scopes, or to understand or check how several agents
share the token. It covers the xero-oauth script, .env
configuration, the multi-agent refresh protocol, Xero app registration,
the refresh-vs-full-flow choice, and the scopes newer tools need.
Upstream @xeroapi/xero-mcp-server is effectively unmaintained, so the
configs run a local build that merges ~28 reviewed upstream PRs plus
local fixes (xero-node v20, token-file locking). Tools and behaviour
therefore differ from upstream's README. For which PRs are in, how to
rebuild after changes, and how to add a new upstream PR, see
references/local-server.md.
The MCP list-* tools paginate at 10 records per page by default; most
accept pageSize (up to 100). Absence on page 1 is not absence in
Xero. Before concluding a record is missing, search by ID directly or
page exhaustively.
See references/searching-records.md
for the rules around list-manual-journals, list-invoices,
list-bank-transactions, and list-contacts.
The schema descriptions for create-manual-journal and
update-manual-journal are misleading on at least two points:
lineAmountTypes: "NO_TAX" is rejected by the Xero API (use the
default — omit the field — which serialises as NoTax).update-manual-journal says "Only works on draft manual journals",
but in practice mutates POSTED MJs (narration, date, lines, status).This makes the void-and-replace pattern (create new POSTED MJ
referencing the old UUID, then update the old to status: VOIDED)
viable via MCP without touching the web UI.
The response strings from the write tools also render input params
verbatim and show line items as [object Object] — so always confirm
state with list-manual-journals manualJournalId=<uuid> after every
write before trusting it.
See references/manual-journals.md
for the full void-and-replace recipe, common error messages, and the
token-expiry coupling that bites multi-step MJ workflows.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Applies local reliability, viewport, tab-safety, stale-ref, widget, and browser-boundary lessons to every agent-browser automation. Use alongside the upstream agent-browser skill whenever navigating, clicking, filling, testing, extracting, taking screenshots, or debugging any website with agent-browser.
中文简介正在准备中。当前显示原文说明。
Size the agent-browser Chromium window to fill its current screen, leaving a configurable bottom margin for the desktop panel. Use before any agent-browser automation that needs the full screen visible without elements overlapping or scrolling off (e.g. Xero, dense web apps).
中文简介正在准备中。当前显示原文说明。
Create Claude Code slash commands, OpenCode command files, Pi prompt templates, and Codex custom prompts that delegate to the right subagent or skill. Use when creating new commands, porting commands across agent platforms, or refactoring existing commands to follow platform conventions.
中文简介正在准备中。当前显示原文说明。
Adds or changes allowed commands in AI agent permission configs for OpenCode and Claude Code. Use when a command needs adding to an agent's allow list, when permission settings for existing commands need changing, when setting up a new AI tool that requires command permissions, or when the user asks to stop being prompted for a particular command.
中文简介正在准备中。当前显示原文说明。
Creates a Beads issue for a described problem or piece of work without doing the work itself. Use when the user invokes `/bc` or `$bc`, or asks to file, log, or record a bead for something. To also carry out the work, use the `bx` skill instead.
中文简介正在准备中。当前显示原文说明。
Stores a persistent memory or learning with `bd remember` so it survives across sessions. Use when the user invokes `/bdr` or `$bdr`, or asks to remember, note, or persist an insight in a Beads workspace.
中文简介正在准备中。当前显示原文说明。