Account and Settings User Guide¶

Settings is BiuMind's hub for accounts and preferences: sign in to your account, plug in your own model API keys, issue long-lived tokens for third-party tools, manage signed-in devices, tweak appearance and per-module defaults, view usage stats, and handle membership subscriptions and version updates — all from one place.

Getting to know the Settings page¶

Click the user area at the bottom of the left sidebar to open Settings. Desktop is a two-pane layout — navigation on the left, content on the right; mobile becomes two stacked pages: the category list first, then full-screen detail once you tap an entry, with the back button returning to the list.

The left navigation is organized by group:

Group	Entries
General	Data statistics, Appearance, Search, My shares, Document processing
Agent	Credentials, Model services, Logged-in devices, API Tokens, Activity, Coding Workbench (desktop), Chat platform
Subscription & billing	Membership Center, Order history, Redemption codes, Referral rewards
Devices	My devices
System	About
Account	Sign out

A few entries (shortcuts, network proxy, data storage, etc.) are marked "soon" — placeholders not yet open, and not clickable for now.

[!NOTE] The Coding Workbench entry appears on desktop only. Logged-in devices manages login sessions, while My devices manages paired biu CLI machines — different purposes, see Multi-device sign-in and device management.

On first use, set up your server address and account under Agent → Credentials:

Server address: your BiuMind server address (the production site or your self-hosted site; http://localhost:8088 for a local dev environment).
Email and Password: your account credentials.
Click Sign in.

After a successful sign-in, session history syncs automatically across your devices; the bottom of the page shows when your access token expires (tokens renew automatically — nothing to manage).

Registering and email verification¶

On the Credentials page click Register, submit an email and password, and the system emails you a 6-digit verification code (if the site has no mail service configured, ask your administrator to get the code from the server logs). Then enter the code on the sign-in page to finish verification — unverified accounts cannot sign in.

Signing out¶

Two entries: Settings → Account → Sign out, or the user area at the bottom of the sidebar. After confirming, local caches are cleared (knowledge base, creation, and subscription caches); coding projects are kept. You land back on the sign-in page.

[!WARNING] Signing out only affects the current device; other devices stay signed in. To take a device offline immediately, use the revoke function in Logged-in devices.

Model services: bring your own API key (BYOK)¶

Agent → Model services lets you plug in API keys from your own model providers (Bring Your Own Key). Once configured, calls to those models go through your key and skip platform billing.

The page shows provider cards on the left and, on the right (wide screens), the configuration detail of the selected provider; on narrow screens, tapping a card opens its config page. Credentials come in two flavors, told apart by the badge on the card:

Badge	Route	Best for
Cloud	Forwarded through the BiuMind cloud to the provider	Works across devices — any client signed into the same account can use it
Desktop	Dials the upstream directly from your desktop client	Upstreams the server can't reach, like a corporate intranet proxy or a gateway you run locally

Supported providers¶

Anthropic Claude, OpenAI GPT, DeepSeek, ByteDance Doubao, Alibaba DashScope, Volcengine Ark (Doubao vision), Google Gemini, Azure OpenAI, Kimi (Moonshot), Alibaba Qwen, Baichuan — plus Custom (any OpenAI-compatible / Anthropic / Google protocol upstream, such as a self-hosted gateway).

Each card shows its credential state:

Not configured (cloud providers show a "Not configured — add a key to skip platform billing" hint);
Configured and verified: a green check at the end of the card, with the subtitle showing the key's last four characters and the upstream address;
Configured but failed verification: a red exclamation mark.

Configuring a cloud credential¶

Click the provider's card on the left.
Enter the API Key (that's the only field for standard providers).
(Optional) Add a note — handy for telling multiple keys apart.
Click Save. When editing an existing credential, leaving the key field blank keeps the current key unchanged.

Configuring a custom provider¶

After selecting Custom, fill in these in addition to the API Key:

Base URL: the upstream address, e.g. https://new-api.example.com;
Protocol: OpenAI-compatible (new-api / vLLM / DeepSeek-style gateways), Anthropic, or Google;
Models covered: declare which models this key covers, comma-separated. Prefix wildcards are supported — glm-* (every model starting with glm) or glm-4.5,glm-4 (exact list).

[!WARNING] Model matching supports only "the full wildcard *" or "prefix wildcards (like glm-*)" — not wildcards in the middle. glm-*-mini is treated as a literal model name and never matches anything.

When a call comes in, the system matches the model name against these declarations; a hit routes to your key instead of platform billing.

Configuring a local direct credential¶

Click Add local direct credential at the bottom of the list.
Base URL is required — e.g. https://intranet-proxy or a provider's official address.
Pick the protocol, fill in the models covered (same rules as above, required), and enter the API Key.
After saving, the card carries the "Desktop" badge, and chat traffic for those models is routed directly from your desktop client to that upstream.

[!NOTE] Local direct credentials currently support the Custom provider only. If you see a legacy direct credential for a standard provider (created by an older version) flagged "legacy dead credential" in red, delete it and recreate it as Custom.

Testing the connection¶

The Test button in the bottom-left corner of the config page runs a connectivity check:

Cloud credentials: tested by the server on your behalf, returning valid / invalid / network and the like;
Local direct credentials: a real request goes out from this machine; on success you'll see "Connected (latency ms) · model name". If the models you entered use a wildcard (like glm-*), it first pulls the upstream's /models list, resolves the first concrete matching model, and tests with that.

Deleting a credential¶

The Revoke (cloud) or Delete (local direct) button in the bottom-right corner of the config page takes effect immediately after confirmation — irreversible; you'd have to re-enter the key if you need it again.

API Tokens (long-lived tokens for third-party tools)¶

Agent → API Tokens issues long-lived Bearer tokens (PATs) for situations where a login session doesn't fit — MCP servers, CI scripts, automation programs. For the full picture see API and third-party integrations.

Creating a token¶

Click Create in the top-right corner.
Fill in a name (like claude-desktop / ci-deploy).
Check the permissions: read (read the knowledge base / search / lists), write (create / edit / merge pages).
Drag the slider to pick the validity: 7 to 365 days, 365 by default.
Click Create; a dialog shows the full token in the clear, in the form bm_<8-char prefix>_<JWT>. Click Copy to save it, then I've saved it to close.

[!WARNING] The secret is shown exactly once, at creation; the server stores no plaintext, and it cannot be viewed again after the dialog closes. Copy it to a safe place immediately.

To use it, put the whole bm_... string into the request header as-is:

Authorization: Bearer bm_xxxxxxxx_eyJhbGciOi...

Revoking a token¶

Each token card in the list shows its name, masked prefix, permissions, last used, and expiry. Click Revoke and confirm.

[!WARNING] The revoke confirmation notes: the current deployment does not enforce revocation, and issued tokens may remain valid until they expire. If you suspect a leak, reissue with a shorter validity and replace every consumer.

One account can be signed in on desktop, mobile, and web at the same time, with data syncing automatically. Each client registers a session under the account on sign-in.

Agent → Logged-in devices lists all your login sessions:

Each card shows the device name (like macOS · MacBook-Pro, Web on macOS), a device-type icon, last active time and IP, and the authorization validity and expiry;
The device you're using carries a purple "Current device" tag and sorts first;
Click Revoke to take that device offline immediately — it must sign in again; revoking the current device is the same as signing out on this machine;
When other devices exist, a Sign out everywhere else button appears at the bottom of the page: everything except the current device gets logged out.

My devices (biu CLI pairing)¶

Devices → My devices manages remote machines paired via biu pair (say, the Mac at home running the biu CLI):

On the target machine, run:

biu pair

The machine displays an 8-digit pairing code; back in the client, click Approve new device in the bottom-right corner and enter the code to confirm.
Once approved, the device gets its device token automatically and appears in the list with its online state (online / offline / last active).

Each device can have its own tool permissions: read-only, read-write within a restricted directory, or full access; and you can revoke any device at any time — its token stops working immediately and it must be paired again.

The biu CLI signs in to the same account via the OAuth flow with biu auth login (a --manual mode is available); see Getting started with the biu CLI.

[!TIP] Logged-in devices manages the sign-in state of each client app / browser; My devices manages the remote execution machines paired through biu pair. Revoking the former means typing your password again; revoking the latter means pairing again.

Appearance¶

General → Appearance has four sections, all applied instantly — no restart needed:

Setting	Content
Interface language	Follow system / 中文 / English
Color theme	18 theme palettes (recommended ones first), with two-tone dots previewing the primary and accent colors — click to switch
Font size	Small / Medium / Large, affecting font size, spacing, and list density, with a live preview card below
Mode	Follow system / Light / Dark

Chat platform¶

Agent → Chat platform configures the out-of-the-box behavior of new chats (for the concepts, see the Chat user guide):

Setting	Content
Default chat mode	Which of Chat / Agent / Task a new chat starts in (Agent out of the box)
Default model (chat mode)	The model new sessions use by default; picking "BiuMind default (unspecified)" follows the platform configuration
Text-to-speech	The voice source for read-aloud: the default is "device-local speech (offline, free)"; pick a cloud voice model and you can further set a voice ID — read-aloud then uses high-quality cloud synthesis, falling back to the local voice on failure
Auto-title from the first prompt	When on, new sessions are named automatically from the first message; when off, the sidebar keeps the "New chat" placeholder until you rename it

Coding Workbench¶

Agent → Coding Workbench (desktop only) configures the executable paths and default working directory of your AI coding agents (for the concepts, see the Coding Workbench user guide):

The Task isolation (worktree) toggle: on, every task runs in its own git worktree + branch (biu/<agent>-<id>), so parallel tasks never interfere; off, all tasks share the working directory and may overwrite each other's files. Switching saves immediately.
Working dir: the directory tasks run in — where agents read and write files.
The three executable paths, biu / Claude / Codex: leave empty to resolve via PATH; use an absolute path for unusual install locations. Each row has a Test button that runs <path> --version and shows the version or an error.
Auto-detect paths & versions: the desktop daemon scans PATH and common install directories (nvm / brew / .local/bin, etc.), auto-fills blank fields, and summarizes each agent's version. The button is disabled while the daemon isn't connected (per-field Test still works).
The bottom of the page shows the effective PATH (parsed from your login shell) — handy for debugging "it works in a terminal but the app can't find it".
Restore defaults empties every path field; Save writes the configuration.

Document processing¶

General → Document processing decides where documents imported into the knowledge base (PDF / DOCX / XLSX / PPTX / EPUB / HTML / MD / TXT) get parsed:

Option	Behavior
Automatic (default)	Small files parsed locally (free; desktop ≤50MB / mobile ≤10MB), large files automatically in the cloud
Prefer local	Parse locally whenever possible (free); only very large files (>200MB / mobile >80MB) go to the cloud
Prefer cloud	Everything is uploaded and parsed in the cloud, billed in credits per page

Local parsing is free; cloud parsing costs credits per page; scanned-document OCR and audio/video transcription are cloud-only. Platforms without local parsing (Windows / Linux) disable "Prefer local" and note that the cloud is always used.

Below that is the Wiki generation model: the model used to auto-generate Wiki pages from uploaded documents. The preference is stored on your account and synced across devices; the default is "follow the platform default", and a model you pick yourself is billed against your usage (your configured API key is used first).

Search¶

General → Search currently has a single toggle: Include notes in unified search. On, the global search page's results include matching notes (title + summary); off, only knowledge base content is searched. It doesn't affect note search inside the Notes module.

My shares¶

General → My shares centrally manages every note-share link you've issued (sharing itself is initiated from the note editor's toolbar or the list's context menu). Each row shows the note title, a status badge (Active / Stopped / Expired / Limit reached), validity, view count (like 3 / 10 views), and whether a password is set.

Available actions:

Action	Effect
Copy link	Active shares only; copies the current share address
Stop sharing	The link stops working immediately; it can be resumed later (same address)
Resume sharing	A stopped share comes back at its original link
Reset link	Issues a new token; the old link dies immediately, and every link already sent out goes dead

Data statistics¶

General → Data statistics is account-level, cross-device usage and activity stats, split into Activity overview and Usage.

Activity overview:

Four metric cards: topics, messages, cumulative tokens, and models used — each with a month-over-month figure (percent change versus last month);
An activity heatmap for the past year (contribution-grid style), with "active N days" and "N-day streak" in the top-right corner;
Two rankings: model usage (messages per model) and topic volume (the sessions with the most messages).

Usage (viewed by month, flip forward and back):

Three cards: today's spend (credits), this month's spend (credits + call count), and number of active models;
A daily bar chart, switchable between credits and tokens;
A spend-by-model ranking;
A call detail table: each call shows model, input / output tokens, TPS, spend (credits), and time — paginated.

AI memory¶

Membership and credits¶

Membership Center¶

Three ways in: the credit badge at the bottom of the sidebar, Settings → Subscription & billing → Membership Center, or the user avatar menu. The page, top to bottom:

Current plan card: plan name, subscription status, and period end date; credit balance (split into permanent and time-limited credits); progress bars for the used portion of your monthly Chat quota and monthly AIGC quota;
Monthly / yearly toggle (yearly carries a discount badge);
Choose a plan: each plan card shows the price (yearly prorated to a monthly figure) and the benefits list (monthly credits, Hub RPM / TPM, daily sandbox tasks, Brain projects); your current plan is marked "Current", and the others show "Upgrade" / "Downgrade" by tier;
One-time credit packs: credits bought once (not affecting your subscription or monthly credits); time-limited packs carry a validity badge (like 30d); click to enter the payment page.

Upgrading, downgrading, and canceling¶

Upgrade: picking a higher plan opens a confirmation showing the prorated difference in detail (credit for the remainder of the old plan, the top-up for the new plan, what you pay now); confirm to proceed to payment;
Downgrade: takes effect at the end of the current period after confirmation;
Cancel subscription: the "Cancel subscription" on the current-plan card offers two options — "Stop at period end" (service continues to the end of the period, no refund) or "Stop immediately + prorated refund";
A canceled subscription can be reinstated with Resume subscription.

Payment¶

On the payment page, pick a payment method, then click Pay now:

Method	Notes
WeChat Pay (QR)	Shows the QR code content and order number; pay by scanning with WeChat
WeChat Pay (H5)	Jumps to WeChat's H5 payment
Alipay (web / mobile)	Opens the browser and jumps to Alipay to pay
International credit card (Stripe)	The overseas card channel

After paying you return to the client; credits usually arrive once the payment callback is confirmed — go back to the Membership Center and refresh to check.

Order history¶

Subscription & billing → Order history lists all orders: amount, payment channel (WeChat Pay / Alipay / Stripe / Apple IAP / Google Play), order type, creation time, and the channel's order number (copyable), with status badges: paid (green) / pending (blue) / refunded (orange) / failed and canceled (red).

Redemption codes¶

Subscription & billing → Redemption codes: enter a code and click Redeem. Four kinds of codes are supported: price-off / discount / credit packs / trial extensions. Common errors are shown with their reasons (invalid code, expired, disabled, already used, not applicable to your plan, currency mismatch, usage limit reached). After a successful redemption, subscription and order data refresh automatically.

Referral rewards¶

Subscription & billing → Referral rewards shows your personal referral code and invite link (one-click copy or the system share sheet), plus referral stats (total invites / rewarded / pending / revoked).

Reward rules:

When an invitee completes their first valid subscription, both sides get 500 credits;
Invites over a threshold from the same IP / device fingerprint within 24 hours count as fraud — no reward;
If the invitee gets a refund or is flagged, rewards already issued are clawed back.

Plan comparison table¶

The client ships a plan comparison table comparing tiers across monthly fee, yearly fee, monthly credits, Hub RPM / TPM, daily sandbox tasks, and Brain projects, with a plan picker at the bottom of each column.

Version updates¶

Update banner: when a new version is detected, a banner appears at the top of the app with the version number and a release-notes summary; Go to download jumps to the official download page. Closing the banner silences it for this session only — it reappears after a restart.
Manual check: the System → About page checks for updates automatically (repeat visits within 60 seconds don't re-request), showing one of three states: "New version found", "Up to date", or "Check failed"; Check again forces a re-query, and when a new version is found the banner carries a "Go to download" button.
Dev builds: a toggle on the About page (labeled "Unsigned · may be unstable · back up your data first"). On, updates come from the nightly build channel instead; a new build triggers "New dev build #N found", with download links matched to your platform automatically (Apple Silicon / Windows / Linux AppImage / Android APK).

The About page also shows the current version and build number (version + build timestamp).

Activity¶

Agent → Activity is your account's event stream: recent operations in reverse-chronological order (PAT created / revoked, with page edits, skill installs, and more to come). Scrolling to the bottom loads earlier records automatically; the top-right corner has a manual refresh.

Chat user guide — chat and agent modes
Knowledge Hub guide — document import and the Wiki
Coding Workbench user guide — the Coding Workbench
API and third-party integrations — PATs and the REST / MCP interfaces
Getting started with the biu CLI — command-line sign-in and pairing