What's New
See what we're building. Every update to BrainRoad — new features, smarter defaults, and the small details that make your AI agent feel like it's truly yours.
Hermes responds faster and handles longer work
The hosted Hermes helper now runs the latest stable runtime, with faster startup, more resilient tools and sessions, richer voice and research support, and better mid-task steering. BrainRoad review still governs anything that would act outside your account.
See context-backed customer work before signup
The new browser-only lead reply test shows how prices, service areas, availability, and approval rules change a customer draft without submitting or saving anything. New buyer guides explain document analysis, small-business assistant workflows, and how Lead Line answers calls from a test-confirmed Sales Call Profile.
Hermes helper upgraded to the latest runtime
The Hermes AI helper now runs the newest upstream release, bringing dashboard session search and filtering, faster session lists, speech-to-text model options, and a stack of chat and reliability fixes from three weeks of upstream development.
Lead Line test calls no longer expire
After approving a Sales Call Profile, owners can call their dedicated number whenever they are ready—there is no five-minute window to race. Every pre-activation call stays review-only until the signed-in owner confirms a successful test, and callers hear a useful answer instead of a carrier application error during normal setup states.
Lead Line tests now follow the current profile
Lead Desk no longer offers to confirm a completed test call from an earlier Sales Call Profile. Historical tests stay visible for review, while every revised profile requires its own fresh production-path test before the line can be turned on.
Sales Call Profile refresh now reads across your business documents
Refresh now considers more of your recent eligible business uploads and Drive files. It keeps strong matches while preventing repeated passages from one file from consuming every evidence slot, giving the receptionist more owner-reviewed facts to answer with instead of falling straight into intake questions.
Lead Line now sounds like a receptionist, not a questionnaire
The receptionist now responds conversationally using only the exact facts and questions in the owner-approved Sales Call Profile. It avoids repeated intake questions, closes gracefully when the caller is done, and lets owners refresh caller-safe answers from newly indexed Business Brain documents before reviewing a new profile version.
Sales Call Profiles now catch missing service answers before testing
Lead Desk now asks owners to confirm that an approved profile includes a useful, public-safe service answer. Sparse approved profiles point directly to creating a stronger version before the next production-path test.
Sales Call Profiles show where every approved fact came from
Lead Desk now links Business Brain facts to their exact source while clearly labeling facts the owner wrote during review. Owner edits no longer appear to have a source document that does not exist.
Lead Line now keeps listening when callers jump in
Callers can interrupt naturally without knocking the receptionist off the call. The live health check now proves that full interruption-and-response exchange, and every caller-facing phone-provider callback has a safe spoken response instead of a carrier application error.
Lead Desk validation keeps keyboard focus where it belongs
Profile validation now keeps focus on the first field that needs attention even while Lead Line status refreshes in the background. Test-call Brain status also uses valid accessible document structure for screen readers.
Lead Line test calls stay safely separate
A test call still exercises the real Lead Line path, but it no longer counts as customer work or prepares a message that could be sent. Lead Desk and Inbox now label the result as a review-only test while real callers keep the normal follow-up workflow.
Lead Line failures now end with a useful answer
Lead Line now keeps its approved AI model binding distinct from the phone carrier, converges captured calls into one lead and review draft, and returns a clear spoken fallback instead of a carrier application error when background processing needs recovery.
Lead Line enters a controlled, entitlement-gated pilot
Eligible pilot accounts can now set up a dedicated sales number from Lead Desk, review a caller-safe Sales Call Profile generated from their Business Brain, test the real call path, and turn it on. Calls capture a durable lead even when the account’s helper is asleep, while follow-up messages still wait in the existing review Inbox.
Team room upgraded to Buzz 0.5, teammates can now chat with your helper
The hosted team room now runs Buzz 0.5, matching the current Buzz Desktop release, and every invited teammate — not just the owner — can talk to the shared BrainRoad teammate in your channels. Approvals are unchanged: anything that leaves the building still waits for a human thumbs-up.
Team room setup accepts the key Buzz actually shows
Setting up your team room now works with the Public key exactly as Buzz Desktop displays it — copy, paste, save. The setup screen also tells you exactly where to find it, and still refuses anything that could be a private key.
Advanced tools now follow the selected helper
Settings no longer opens an OpenClaw-only browser from a Hermes helper. Unsupported and temporarily unavailable tools stay keyboard-reachable, explain what is needed, and block the incompatible navigation.
Developer Triggers load reliably and recover honestly
Settings can read the selected helper’s live webhook URLs and token again. If that secure read is temporarily unavailable, the page now explains the problem and offers a guarded Retry instead of showing a loading placeholder forever.
The Inbox now keeps complete, recoverable review history
Pending work includes due snoozed items, Recent includes every approval and expiration, and older cards load in stable pages without hiding search results. Failed queue, older-page, item, thread, defaults, and automation reads show an in-place Retry; failed declines keep your reason; email drafts can add Bcc; review, snooze, and settings controls prevent repeated actions while preserving keyboard focus; and critical finding labels retain readable contrast.
Runs, usage, and proof now fail closed and keep the Inbox authoritative
Run output stays attached to the selected helper, incomplete streams report an error, and proposed outside actions open the exact Inbox card for full review instead of approving from a summary. Work Ledger filters, bounded cursor history, briefings, native usage links, and diagnostic counts now distinguish outages from empty results, preserve live events, guard repeated actions, keep keyboard focus on the result, and make the Activity descriptions easier to read.
Uploads, questions, and connected sources recover without losing your place
Meeting notes, file retries, suggested collections, and Brain questions now accept one action at a time while keeping the active control keyboard-reachable. Connected-source outages show a real retry instead of looking unconfigured, Google Drive folder reads can recover in place, and removing imported Drive files or Calendar events now requires explicit confirmation.
Business Brain actions keep their place and show where the facts came from
Collection changes, file retries, downloads, health checks, wiki builds, and Beacons now stay keyboard-reachable while they run and move focus to the result when a control disappears. Search failures can retry the same query, receipt source links use correct document structure, and opening an original file reliably starts from the click instead of being lost to a popup blocker. Delegated reviewers see only shared read controls, Beacon outcomes cannot race across tabs, and BrainRoad waits for your saved-key and included-credit status before deciding whether setup is needed.
Helper comparison now matches current runtime boundaries
The OpenClaw and Hermes guide now explains what both helpers share, which dashboard surface is OpenClaw-specific, how migration works, and exactly where BrainRoad review applies. Retired Company-workspace language and blanket approval claims are gone.
Keyboard navigation is clearer across the dashboard
Every dashboard page now exposes a reliable page heading and moves keyboard focus there after navigation, including Console and Browser. Mobile navigation and admin dialogs now trap and restore focus correctly, helper lifecycle actions keep their keyboard position while requests are pending, and horizontally scrollable code examples are keyboard-accessible. Hermes chat is no longer hidden behind OpenClaw-managed key state, and every passive Business Brain read failure now offers an in-place Retry. File upload, admin reports, danger controls, billing links, and incomplete Team invitations now have clearer labels, contrast, focus behavior, and recovery guidance.
Pilot workspaces start with vertical setup worksheets
Founder-assisted wedding-venue and IT-service pilots can now begin with four clearly marked worksheets for service rules, quotes or discovery, FAQs, and a Business Interview. They are prompts to complete with the owner—not claims about the business—and real activation still requires real documents, inbound work, and reviewed actions.
One honest OpenClaw security library replaces 19 overlapping posts
Five maintained guides now cover managed-host pricing, self-hosting, gateway hardening, skill vetting, and BrainRoad’s exact isolation and approval boundaries. Fourteen overlapping pages redirect to the closest answer, and old claims about public-by-default gateways, universal approval coverage, and invented customer stories are gone.
Team-room setup now follows the upstream Buzz app
Copy your public identity from Buzz Desktop, let BrainRoad host the relay, then paste the relay address into Buzz and add the BrainRoad teammate to a channel. There is no custom Windows link and your private key never leaves Buzz; a stopped room can still recover from a mistyped or lost owner identity.
A shared Team room replaces the Company workspace
The dashboard's Company workspace is now Team: one shared room per account, built on the open-source Buzz project, where the people on your account and your AI helper work from the same Business Brain. Approvals still run through your Inbox — the room is not a second queue. Joining currently needs the Buzz desktop app from the upstream project, which we do not distribute, and the Team page says so plainly.
Turn your reviewed replies into suggested Brain guidance
BrainRoad can now notice a reply you substantially rewrote, a question you approved repeatedly, or a pattern you declined more than once and suggest a reusable template, FAQ, or business rule. The suggestion stays in your review Inbox, shows the decisions it came from, and changes the shared Business Brain only when the account owner approves it.
Add meeting notes with the date they happened
Paste meeting notes or upload one transcript with its real meeting date and title. BrainRoad makes it searchable with a clear meeting-date source label and can match a same-title Google Calendar event, without recording calls, transcribing audio, or changing the calendar.
Documentation rewritten to match the product you actually use
The docs section now covers the Business Brain and the review Inbox with dedicated guides, documents both OpenClaw and Hermes helpers, and reflects current pricing, developer access, and channel behavior. Pages describing retired or never-shipped features were removed, with old links redirected to their closest live replacement.
Brain wiki rebuilds reliably after a failed compile
The "Build compiled wiki" button now always queues a fresh compile. Previously, if a compile had ever failed, later rebuild requests were silently ignored, and the setup-complete email after your first Brain upload could be blocked by the same class of bug.
The AI agent platform guide now shows the current product
The guide now renders the same governed lead-response, hosted OpenClaw and Hermes capability, and delegated-review story as the rest of the site instead of an older hard-coded platform page.
Invite a teammate to help review the work
Account owners can now invite one delegated reviewer into the shared Business Brain and Triage queue. Every edit, approval, decline, and send grows an attributed account work record; teammate approvals execute, while billing, helper controls, Brain changes, and the authority to widen autonomy stay with the owner.
A clearer path from real business facts to a reviewed lead reply
The site, pricing, onboarding emails, and dashboard setup now teach one complete workflow: add a price or policy source, forward a lead, inspect the grounded draft, and review the server-side send record. Capability wording now preserves the full OpenClaw and Hermes toolset while distinguishing managed email, optional text-only SMS, owner-configured integrations, and the authority boundary for each path.
Drafted-reply alerts are one-time and pressure-free
When a real customer reply waits in Triage for more than a day, BrainRoad sends one neutral notification. Leaving it for later produces no follow-up reminders; a genuinely new draft can still notify you once.
Natural questions now find obvious calendar and file matches
You can ask about an appointment by its title or describe a Drive record in your own words without copying exact timestamps or phrases. If the optional meaning-based search is slow, BrainRoad now falls back quickly to its indexed words and titles instead of making the whole answer wait.
Google Calendar checks now report the truth
A completed calendar import now clears an old error, records the successful check, and shows what was added, updated, or removed. If a later check fails, events already in your Brain stay available and the card shows when the failed attempt happened.
File problems now show the next useful step
The Upload tab, Library, and document reader now identify the affected file, distinguish reading and indexing failures, show when a retry is underway, and include a safe support reference. Brain totals and health checks use the same ready, processing, failed, and Knowledge-page definitions.
Background Brain reading works with Anthropic
Knowledge pages, receipt details, health checks, and needs-attention scans no longer mistake a valid Anthropic setup for a missing page. BrainRoad still keeps its platform model reserved for background reading rather than spending it on interactive questions.
Failed file processing no longer looks nearly finished
A file that cannot be read now moves immediately to a failed state instead of lingering at 97%. Brain health no longer treats a partial wiki as complete, and receipt facts can use BrainRoad’s background model without exposing that key to interactive questions.
Receipt photos are read reliably and repeated uploads stay visible
Phone photos now use an image reader that is included with BrainRoad instead of depending on a model download during processing. Uploading the same receipt again keeps both records and marks the newer one as a possible duplicate.
Your Business Brain can remember what is on your Google Calendar
Connect your primary calendar and BrainRoad will bring in appointments, participants, and renewal dates as read-only business context. Upcoming items can appear in “needs attention,” with links back to the source record; BrainRoad cannot create or change events or invite anyone.
Receipts can fill your Business Brain from email or a phone photo
Forward a receipt or invoice attachment, or take a photo from the Upload tab. BrainRoad reads the vendor, date, total, and tax with links back to the source, and points out possible duplicates without deleting either record. It is reference-only and never posts to bookkeeping software.
Your Business Brain can keep up with a Google Drive folder
Connect one folder and BrainRoad will read its existing files, then keep new, changed, and removed files in sync automatically. You choose whether imported files stay or leave your Brain when you disconnect, and BrainRoad never edits anything in Google Drive.
Answers now show the records behind them
Every answer now says how many business records it checked and lets you open a clear source list with the exact supporting excerpt. BrainRoad also warns you when an answer relies on a record that has been replaced by a newer one.
Answers point to the exact page, with a record for each customer and vendor
Source links now open the exact page or paragraph behind an answer and highlight the supporting words. New read-only pages gather what your Brain knows about each customer, vendor, property, or organization—with facts, history, disagreements, and related records tied back to their sources.
Search results now take you straight to the answer
Searching your Brain now opens results directly: documents open in the new reader, Knowledge pages open to the right page, and past questions land in Ask ready to re-run. You can also view an uploaded original right in your browser, and build your Knowledge pages on demand if they have not compiled yet.
Lead reply checks recover when a model fails
Managed helpers now switch to a backup model when their primary model is unavailable. Unconfigured chat integrations no longer appear connected or keep a helper running, while real connected channels stay online.
Read everything in your Business Brain
Every document in your Brain library now opens to a full readable view — including the notes your helper writes — with a clear "Written by your helper" badge, the original file available for download, and delete right there. A new Knowledge tab shows the pages your Brain compiles from your documents, with every claim linked back to its source.
Opening your AI company works again
The Company tab briefly failed with a setup error when launching a fresh Paperclip workspace after a platform upgrade. First-time setup now completes and your CEO agent arrives with its full onboarding playbook already in place.
Helper workspace icons and offline support restored
After the latest helper runtime upgrade, some workspace icons showed as broken images and the page address could drift to a link that failed on reload. The workspace now keeps its correct address, so icons, your helper’s avatar, and offline support all load reliably again.
A clearer, more reliable Lead Desk
You can now test your lead desk before a real customer writes, see when your helper is waiting for capacity, and recover an email draft that needs attention. Every draft still waits in Triage for your review, and nothing sends without your approval.
Helpers with connected chat channels stay online
If your helper is connected to Telegram, Discord, Slack, or another chat channel, it now stays running around the clock so messages are never missed while it saves capacity. A message from you also keeps it awake, and helpers without channels stay up longer between uses.
Sleeping helpers no longer miss customer emails
Agents that pause to save capacity now wake automatically when you open their workspace or a customer email arrives. Inbound messages wait safely until a reply draft is ready in Triage, so a sleeping helper never drops a lead.
Your helper now keeps a clear Work Ledger
Activity now opens to an owner-friendly record of pending work, what you changed or approved, what actually happened, and how trust rules changed. Usage shows tokens even when dollar cost is unavailable, while technical logs stay in their own tab.
Upload a file, then ask it directly
Brain uploads now recover cleanly from slow processing and offer an immediate Ask about this file handoff when indexing finishes. The question stays tied to the exact file, and a failed request keeps a safe retry path until you receive a cited answer.
Your review edits survive a newer draft
If your helper updates a card while you are editing it, BrainRoad now keeps your changed fields and brings in the untouched parts of the newer draft. You also see the exact customer message before approval and can safely retry a provider rejection that happened before delivery.
Save booking defaults and send real invitations
Set your business time zone, currency, quote validity, appointment length, and usual location once. Approved bookings now include a real calendar invitation with a stable identity, so customer calendars show the right time and a changed booking does not overwrite an unrelated event.
Connecting another AI no longer takes guesswork
New Brain and Inbox connections start with narrow permissions, bind to the helper you selected, and include copy-ready Claude Code and Codex commands. A built-in connection test proves the Brain and approval queue are reachable before you leave the page. If a key needs attention, its next-step link now lands on Developer Access instead of a dead hostname.
Edit every price and booking before it goes out
Quote cards now let you change line items, quantities, prices, notes, and the valid-until date before sending. Booking cards let you correct the title, local time, time zone, duration, and location, and both the review card and customer email show the same named time zone.
Connect Claude Code or Codex to your Business Brain
Create a scoped API key and Claude Code, Codex, or another MCP client that accepts bearer headers can search your Brain, ask it questions, and propose work into your approval queue. Approving stays a human decision in your dashboard. Hosted ChatGPT and Claude.ai connections will follow through OAuth, not pasted API keys.
Your agent now drafts quotes and calendar bookings
Two new card types land in your approval queue: a quote with line items and a total, and a calendar booking with the date, time, and place. Your agent drafts them from the conversation; you review before anything reaches the customer — same as every email.
See how complete your Brain is at a glance
A new completeness meter shows how much your Brain actually knows about your business and what's still missing, so you know what to upload next. Live agent activity is back on the dashboard feed too, so you can watch work happen as it happens.
Tighter walls around what your agent can reach
We closed a path where an agent's mailbox access could have been used to send email without going through your approval queue — agent pods can now only read mail, never send it directly. External connections also moved to a curated catalog instead of arbitrary URLs. Review-before-send is enforced by the network, not just the rules.
New trial agents start on a stronger default model
Agents running on the included OpenRouter trial key now default to Grok 4.5, which handles tool calls and multi-step work more reliably. Your own model picks are untouched.
Coming back after your trial is now easy — and 15% off
If your trial lapses, you'll get a short series of emails reminding you what your Brain still holds and exactly when your workspace would be archived — with 15% off your first month applied automatically at checkout. Lifecycle emails now say precisely what happens to your data and when, and billing recovery no longer wakes an agent you deliberately stopped.
Checkout always shows monthly and annual side by side
Some upgrade buttons were sending people straight to an annual-only checkout. Every upgrade surface now leads with monthly and shows the annual option explicitly, so the plan you pick is the plan you meant to pick.
All three agent runtimes updated to their latest releases
OpenClaw, Hermes, and Paperclip all jumped to their newest upstream versions in one coordinated update — newer models, faster tool calls, and upstream fixes across chat, channels, and the AI company layer.
Your agent consults a Context Brief before answering big questions
Ask your agent "what do you know about my business?" and it now starts from a structured brief the Brain assembles — your real business documents by name, the key facts, and what's still missing — instead of improvising from raw search. The Brain also tells your actual business context apart from setup and demo material.
The whole product now leads with your Business Brain
The website and dashboard were rewritten around what BrainRoad actually is: a private Business Brain your AI helper works from — files, customer history, FAQs, quote templates, service rules — with review before anything goes out. Upgrade prompts and Brain-health guidance now point the same direction.
Your agent now texts back, reads attachments, and follows up
Agents can reply to SMS through the same approval queue as email. Email attachments (contracts, spreadsheets, PDFs) are saved into your Brain automatically. Quiet threads get a follow-up suggestion after a few days, snooze has real duration choices, and the thread view shows the whole conversation.
Email threads now handle the whole group
Your agent tracks everyone on an email thread — not just the last sender. Drafted replies default to reply-all, you can edit the Cc line before approving, and the thread view shows every participant with how they joined the conversation. Auto-approved replies stay locked to people already on the thread.
Your two agents now work together out of the box
New accounts get the OpenClaw ↔ Hermes bridge turned on from day one, so your two agents can hand work to each other without setup. The Settings toggle stays as the off-switch, and device pairing now heals itself if it ever breaks — no more manual re-pairing.
A live QA sweep cleaned up eight dashboard rough edges
We ran a full session on real hosted accounts and fixed everything we hit — from workboard pages erroring on a stopped runtime to sturdier gateway defaults. The dashboard should feel noticeably less fragile in the corners.
Agents can call outside services — with your approval
Connector calls are a new kind of approval card: your agent proposes a call to an external service, you see exactly what it wants to reach, and nothing executes until you approve. The call itself runs on our servers with your credentials held there — never inside the agent pod.
Drive your hosted Hermes agent from the desktop app
Grab a Remote URL and device token from your dashboard, paste them into the native Hermes Desktop app, and it connects straight to your always-on hosted agent. Your Brain, approvals, and identity stay server-side; you manage every connected device from a registry you control.
The Inbox 'Automation' tab is now 'Approvals'
Renamed to say what it actually is: the place where you approve your agent's outgoing work. Scheduled tasks from your agent runtime now show up alongside it, so recurring work is visible in the same place.
Trial keys now heal themselves
If your included OpenRouter trial key ever got orphaned — say, after a reset — your agent would fail with a cryptic auth error until someone noticed. A background sweeper now detects and re-issues broken trial keys automatically.
Your assistant can earn the right to auto-send
After you approve a handful of routine replies in the same email thread, BrainRoad now offers to handle that thread for you, capped per day and revocable in one click. High-stakes messages still come back for review, and any decline or misfire pauses it automatically. Less rubber-stamping, more time back.
Ask your business documents anything
Upload contracts, leases, financials, and sales docs, including scanned PDFs, and ask plain-language questions. Your assistant answers with the specifics and points to the exact document it pulled from, so the business stops living only in your head.
Agent chat loads cleanly from agent pages
Agent chat now opens the selected runtime chat cleanly from the Agent page. The result is boring in the best way: click Chat, land in the conversation, and start talking to the agent.
Agent chat now opens the native runtime
The main chat button opens the selected runtime chat. OpenClaw and Hermes dashboards stay one click away for configuration, logs, and advanced controls.
Console feels like SSH — GPU-accelerated and searchable
The dashboard terminal now renders on the GPU (WebGL), so scrolling, htop, and build logs feel snappy. Press Ctrl-F (Cmd-F on Mac) to search the buffer with an inline find bar. Bursty output is batched into a single write per frame, so the terminal stays responsive even when an agent is dumping thousands of lines at once.
Console tells you when another tab already has it
Open the console in a second browser tab and you now get a clear "this is already open elsewhere" banner with a one-click Take Over button — no more silent reconnect cycles fighting over the same shell. Behind the scenes, bash readline shortcuts like Ctrl-B and Esc-driven TUIs (vim, Claude Code, Hermes) feel snappy again.
Console sessions survive agent switching
Switch from OpenClaw to Hermes and back without losing the shell you were using. Each runtime now keeps its own persistent console session, Hermes keeps its native terminal colors and width, and the reset button gives you an explicit fresh start when you want one.
You always know which agent you’re operating on
Every dashboard page now shows a small bar at the top with the active agent's name, product (Hermes or OpenClaw), and status — one click to switch. Hop into Brain, Activity, the terminal, anywhere, and you never lose track of whose data you’re looking at. Settings dropped its second switcher; the one in the shell is the source of truth now.
Open the terminal in one click on either agent
OpenClaw users now get the same prominent “Open OpenClaw shell” button next to “Open chat” that Hermes already had — no more hunting for a small footnote. Every agent-scoped link in the dashboard also carries the active agent through the URL, so two browser tabs on two different agents can never send you to the wrong shell.
Hermes terminals coexist safely across tabs
Opening Hermes in two browser tabs used to silently kill the first session. Now the second tab tells you politely that another Hermes shell is open, and when it disconnects the next tab starts right away. The terminal also fits the viewport cleanly and disconnects without leaving a stale "connected" indicator.
Hermes now has a hosted OpenAI-compatible endpoint
Hermes agents now expose a BrainRoad-hosted `/v1` endpoint for external clients, secured by the same gateway token as the dashboard. Dangerous Hermes command approvals can also appear in your Triage inbox and resolve the live blocked session from there.
Hermes chat now opens as a real conversation
The Hermes dashboard chat page now starts with a modern message thread instead of dropping you into a terminal. Messages, reasoning, tool calls, and native Hermes approvals stream inline, with the terminal still available as a fallback tab.
Your AI agent gets to work the moment you sign up
A brand-new BrainRoad account now arrives with a starter document already in your Brain and a sample Triage card waiting for your approval — so you see how the approval path feels before you have to set anything up. Channels stay where they belong, inside your agent's own UI, when you're ready for them on day two.
Choose your AI runtime — Hermes joins OpenClaw
Every new BrainRoad account now ships with both OpenClaw and Hermes Agent pre-installed (and stopped). Start whichever fits how you like to work, run them side by side, or pick the one whose answers you like better. Same Brain underneath either way — your documents, notes, and memories travel with you.
A simpler dashboard with five destinations
The sidebar collapsed from a long advanced-feature list to five primary destinations — Agent, Inbox, Brain, Company, and Settings. The deeper surfaces (Billing, Usage, API, Activity, Console, Browser, Proof) all still live at their old URLs and are one click away from Settings. Less to scan, faster to find what you need.
Connect any messaging channel from inside your agent
Adding WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Matrix, and the rest now opens your agent's native channel manager instead of a four-channel BrainRoad picker. You always see the full, current list of supported channels — and per-channel credentials are managed where they actually live.
Billing now self-heals when Stripe drifts out of sync
If your account ever got disconnected from Stripe — say, after deleting and re-signing-up — the recovery path could leave upgrades or trial restoration stuck. Billing now verifies Stripe ownership before relinking and reconciles automatically when state diverges. Fewer support tickets, more clean upgrades.
Mobile homepage hero no longer clips on smaller screens
On narrower phones the hero section was cutting off the headline before you could read the first line. Re-tuned the layout so the full message lands cleanly on every screen size — no scrolling required to know what BrainRoad does.
Onboarding emails now follow what you actually do
Your getting-started emails used to fire on a fixed signup timer — sometimes before your agent was even ready, sometimes long after you'd already moved on. They now key off real activation moments: agent provisioned, first prompt sent, first reply received. Better timing, fewer misses, less inbox noise.
Your first week now comes with a built-in job guide
New users get a first-week playbook in the dashboard and onboarding flow — a clear, day-by-day path through what to set up, what to try, and what your agent should be doing for you by the end of the week. You don't just have an agent running; you have a plan to put it to work.
First-conversation reliability across flaky networks
If your network briefly dropped or the gateway stalled while you were sending your very first message, the chat could end up half-paired and stay there. We now retry pairing with explicit approval, hold onto connect challenges through the handshake, and re-run transient first turns automatically. Your first conversation lands every time.
Reuploading a Brain document is now obvious
Drop the same file into your Brain a second time and the Library now shows it clearly as a re-upload — replacing the old version, not silently adding a duplicate. Less guessing whether the new version actually took.
Your Brain has a Context tab — see how everything connects
Open your Brain and you'll find a new Context view that maps the entities and relationships across all your documents. People, projects, deadlines, contracts — see them as a graph of how your work actually fits together. Your agent reads from the same map, so its answers already understand the connections.
Brain runs on managed Postgres now — faster and more durable
Moved Brain document search out of per-user files and into our managed Postgres database. Indexing finishes faster, lookups feel snappier, and your Brain data now rides along with our nightly backups instead of living on a separate volume. Nothing new to learn — uploads just finish sooner.
New trial agents start on a smarter default model
We swapped the default model new agents boot with on the included OpenRouter trial key. The new pick (Kimi K2.6) handles tool calls more reliably than the old default — so trial agents stop inventing reasons they "can't" do something and just do it. Your existing model picks are untouched.
Report a bug and What's new — one click from your avatar
Click your account avatar and you'll see two new shortcuts. Report a bug opens a pre-filled email with your user ID, agent ID, current page, and browser info already attached — no debug-info hunting. What's new jumps straight to this page. Less digging, faster help.
More room on every page — desktop top bar is gone
We removed the desktop top bar and folded your account into the sidebar, where it's always one hover away. The result: 52 extra pixels of vertical space on every dashboard page. Your status card, activity feed, and chat all get more room to breathe.
Your dashboard loads faster and tells you what it's doing
Opening the Agent tab no longer leaves you staring at a blank page. You now see a clear "checking your agent status" cue right away, and switching tabs and coming back is instant instead of triggering a full reload. Small wait, bigger difference.
Your dashboard now matches how your AI works
BrainRoad now explains your agent more like a personal operator with its own work identity, so the front door and dashboard make the product clearer. The Inbox, identity copy, and advanced internal labels now point you toward what matters first and tuck infrastructure behind the scenes.
Blog posts now take you straight to a live example
Every role-focused article on brainroad.com — receptionist, answering service, email assistant, executive assistant — now links you directly to the matching live page where you can see exactly how your agent would handle that job. Read about it, then try it. No more hunting for the right product page after you're already interested.
Your Brain reads more reliably now
The pipeline that ingests your contracts, emails, and notes into the Brain was occasionally stalling on large or complex uploads. We tightened how the ingestion worker talks to our database so files process end-to-end without surprises. Drop more in, worry less.
Role pages rewritten around Business Brain workflows
Our public pages for each role — AI receptionist, answering service, email assistant, executive assistant — now lead with the truth: your Business Brain carries the context, the AI helper drafts the work, and review keeps customer-facing actions safe. Claims we couldn't stand behind are out. What you see before you sign up is what you get.
Delete your account and come back — the trial now resets cleanly
If you signed up, deleted your account, and then signed up again, the trial and billing state used to get tangled between Stripe and our AI company layer. Cleanup is now airtight on both sides, so a fresh signup always gives you a fresh 30-day trial — no leftover state, no confusing errors.
Billing stopped throwing surprise errors
A few edge cases in the billing API were returning 500 errors when users checked their plan or upgraded. Those paths are now hardened and covered by new tests. Fewer dead-ends when you're trying to give us money.
Your agents now have a Brain
Drop your contracts, financials, meeting notes, and emails into the new Brain tab. Every agent you hire instantly searches them, cites them, and writes back what it learns. The brain grows with you — uploads, inbound emails, SMS, and the work your AI company does all become part of the same searchable, grounded memory.
Beacons — your Brain speaks up before you ask
The Brain now surfaces what matters on its own. When it reads something in your documents that connects to a decision, a deadline, or an open question, a Beacon appears — a short, cited finding you can act on without running a single search. Less digging, more signal.
Email replies now stay on the same thread
When your agent emails someone from its managed `@mg.brainroad.com` mailbox, replies now come back attached to the same conversation instead of starting over from zero. The thread keeps the reply headers and continuity your agent needs to follow through without losing context.
See when your plan can add another agent
If your plan supports more than one agent, the dashboard now shows the option to add another one right from the agent switcher instead of hiding it. You can also see the real plan limit coming back from billing, so expanding from one agent to a second or third feels explicit instead of guesswork.
Upgrade prompts now keep your annual pricing choice
If you prefer annual billing, BrainRoad now carries that choice through the inline upgrade prompts across your dashboard instead of snapping back to monthly pricing. Trial nudges, near-expiry reminders, and upgrade buttons now stay consistent with the plan you were already considering.
When free credits run out, Key Concierge picks up the handoff
If your trial key budget is exhausted but your key state is still present, BrainRoad now sends you straight back into the guided Key Concierge flow instead of leaving you with a warning and no next step. You get a clearer recovery path for bringing your own key and getting your agent moving again.
MCP Tools Explorer — see everything your agent can do
The new API page shows all 110+ methods your agent supports, organized by category — messages, channels, cron, files, and more. Browse the full tool list, copy a ready-made MCP config for Claude Code or Cursor, or set up local access with one command. If your agent can do it, you can find it here.
4 new case studies showing real agent workflows
Published four in-depth case studies covering how people use BrainRoad agents for solo research, freelance output scaling, ops management, and on-call engineering. Each one breaks down the workflow, the tools involved, and the measurable results.
Console and Browser go full-screen
Both the Console and Browser tabs now use the full window — edge-to-edge, no wasted space. The Console has an integrated toolbar with connection status and quick actions built right in. It feels like a real terminal, not a widget.
Celebration screen when you subscribe
Subscribe to Pro and you get a proper welcome — confetti, a thank-you message, and clear next steps. Small detail, but it matters. You just made a commitment and the product should acknowledge it.
AI Company page redesigned with guided onboarding
The Company page now explains what the AI company is, shows three use-case cards (research team, content pipeline, customer service), and walks you through getting started. No more blank page if you haven't set it up yet.
Trial credit increased to $5 — plus 15% off for early adopters
Every new agent now gets $5 in OpenRouter trial credit (up from $2), so you can have real conversations before committing. And if you upgrade on monthly billing, 15% off your first month is applied automatically at checkout.
Smart upgrade reminders that respect your timeline
As your trial progresses, you'll see gentle nudges at three key moments — when you're getting value, when your trial is halfway through, and a few days before it ends. Each one shows your actual usage stats so the decision is informed, not pressured.
Error tracking and session replay — we see what breaks before you report it
Added Sentry error tracking across the entire stack and Microsoft Clarity session replay on the dashboard. When something goes wrong, we see the exact error, the user session, and the steps that led to it. Bugs get fixed faster because we catch them first.
Side-by-side comparisons: BrainRoad vs Lindy, Manus, and Relevance AI
Three new comparison pages break down how BrainRoad stacks up against Lindy AI, Manus AI, and Relevance AI on pricing, features, and approach. Honest assessments — we tell you where each platform does well and where BrainRoad is the better fit.
Gateway stability improvements — more memory, fewer restarts
Increased memory allocation across all gateway tiers to handle the latest OpenClaw releases. Agents that were occasionally restarting under heavy workloads now run stable. Also added automated OOM detection alerts so we catch issues before they affect you.
Agent troubleshooting guide
A new help article walks you through the most common agent issues — connection drops, slow starts, API key errors — with step-by-step fixes. Written so you can solve problems yourself without waiting for support.
Your agent arrives alive — named, emailed, and ready to talk
New agents no longer boot to a blank setup screen. Sign up and your agent appears immediately with a name, a dedicated email address, and a live chat window. The old "Launch Agent" button and multi-step wizard are gone. You go from signup to conversation in under a minute.
Rename your agent from the dashboard
Click your agent's name to give it a new one. The change updates everywhere — the dashboard, the agent's email address, and the identity it uses when it introduces itself. Your agent is yours to personalize.
8 new blog posts covering the full agent experience
Published guides on API key setup, hiring AI team members, building research teams, automating content pipelines, running AI customer service, troubleshooting, and the activation bottleneck most people hit. All written to help you get more out of your agent.
AI Key Concierge — guided setup replaces the blank key form
Setting up your API key used to mean staring at a dropdown and figuring out provider jargon on your own. Now you get three clear paths: try free with one click, bring your own key with step-by-step instructions, or let us recommend one based on what matters to you. After setup, you go straight into your first conversation — no dead dashboard.
We now track every activation step — so we can fix what slows you down
BrainRoad now measures every step of your journey — from signup to provisioning, API key setup, first message, and first return visit. When users get stuck anywhere in that flow, we'll see it immediately and ship a fix. The activation experience is now on continuous improvement.
Your dashboard tells you exactly what to set up first
If you sign up and skip the API key step, your dashboard now shows a clear activation prompt right above the Getting Started checklist — and the Start Chatting button stays locked until your key is in place. No more cryptic errors when you try to chat. One clear action, and you're unblocked.
Your account now includes a full AI company
Every BrainRoad agent now comes with Paperclip — an AI company layer for free. Your agent can hire specialists, delegate tasks, and manage a team autonomously, while you stay in control with approval flows. It's all built in and ready to use from your dashboard.
Your agent's infrastructure fixes itself
We deployed intelligent runbooks that detect and automatically fix common infrastructure issues — network glitches, Pod scheduling misses, volume attachment timeouts. When something goes wrong, alert rules trigger automated remediation before you notice. Your agent stays available and responsive without your intervention.
Smart defaults for lower costs
Your new agents now default to Claude Sonnet (75% cheaper than Opus) and use smart heartbeat scheduling to avoid unnecessary idle workload. These defaults cut costs dramatically without sacrificing capability. Switch to Opus for premium tasks any time from the dashboard.
New features now get a blog post within 48 hours of shipping
We wired up our AI company to the content pipeline: when engineering ships a feature, a content task is automatically assigned to our AI CMO, who writes the announcement and updates this changelog. This post was written by Niobe, our AI CMO — triggered automatically when the feature was marked done. Every feature you care about gets a write-up fast.
Run multiple agents — one for work, one for personal, one for your side project
Pro users can now provision up to 3 separate AI agents, each with its own API keys, skills, email address, and configuration. Spin up a dedicated agent for each context in your life — no more sharing one agent across everything. Add a second or third agent from the dashboard in seconds.
Smart sleep — your agent rests when idle, wakes when you return
Inactive agents now auto-pause to save resources. Visit your dashboard and your agent wakes up in about 30 seconds — all settings, API keys, and data preserved. Pro subscribers are never paused. Free-tier agents sleep after 4 hours of inactivity; configured agents after 7 days with no autonomous activity.
Build an AI company — org charts, tasks, and budgets for your agent team
Your agent can now hire its own team. The new AI Company dashboard lets you create a company of AI agents with org charts, task management, cost tracking, and board governance. Start with your existing agent, add specialists, and let them coordinate work autonomously. Click "AI Company" in the sidebar to get started.
Simpler sidebar — less clutter, more focus
Removed the Scheduled and Skills pages from the sidebar. Both are available in the OpenClaw dashboard, and having them in two places was confusing. The sidebar is now clean: Agent, AI Company, Console, Browser, Activity, Usage, API, Settings, Billing. Every item earns its spot.
Dashboard loads instantly — no more polling
Your agent now pushes its own status to the dashboard instead of the dashboard asking every few seconds. The result: your agent page loads instantly, status updates appear in real time, and the API does less work. A small change you'll feel every time you open the dashboard.
Your agent has its own email address and phone number
Every agent gets a dedicated email address automatically. Pro users can assign a real phone number too. Your agent can send emails, receive texts, and handle 2FA verification codes for the services it signs up for. It has a real identity — not just a chat window.
People can email and text your agent directly
Your agent's email and phone number aren't just for sending — they receive too. When someone replies to an email your agent sent, or texts its number, the message arrives in the agent's activity feed. Your agent can carry on full conversations over email and SMS without you in control.
Your agent can log into services with 2FA codes
Store TOTP secrets for any service and your agent generates its own 6-digit login codes on demand. GitHub, AWS, Slack — anything that uses an authenticator app. Your agent handles its own security, just like an employee would.
Give your agent access to Git and GitHub
Your agent can now have its own SSH keys and GitHub personal access tokens. It can clone repos, push commits, and open pull requests under its own identity. Credentials are encrypted at rest in a secure vault — you add them once, your agent uses them whenever it needs to.
Encrypted credential vault for agent secrets
API keys, passwords, tokens — your agent needs credentials to do real work. The new credential vault encrypts everything with AES-256-GCM before it touches disk. Each agent gets its own isolated vault. You manage credentials from the dashboard; your agent accesses them automatically.
Connect Claude Code, Cursor, or any MCP client to your agent
The new Developer Access page gives you everything you need to integrate your agent with external tools. Generate API keys for programmatic control, copy a ready-made MCP config into Claude Code or Cursor, and your agent becomes a tool that any AI assistant can call — send messages, manage channels, run cron jobs, and access all 90+ methods.
Control your agent from scripts and CI/CD
A new public REST API lets you start, stop, and check your agent from anywhere — shell scripts, GitHub Actions, Zapier, or your own code. Authenticate with API keys you create from the dashboard, and subscribe to real-time events via Server-Sent Events to react the moment something happens.
Smarter onboarding that tracks real progress
The channel setup flow now watches your agent's actual connection status instead of relying on checkboxes. If your WhatsApp or Telegram link drops, the dashboard reflects it immediately. The "I'm Done" button works reliably, and you won't see the setup screen again after completing it.
Security hardening across the stack
We completed a full-stack security review and shipped 12 fixes — tighter input validation, stricter CORS, improved token handling, and hardened Kubernetes network policies. Your agent's infrastructure is more secure without any changes on your end.
Token usage dashboard with cost tracking
See exactly how many tokens your agent uses and what it costs. The new Usage page breaks down input, output, and cache tokens by day or by model — with estimated costs based on each provider's pricing. Filter by time range to spot trends and keep your API spend under control.
Redesigned agent page and new Settings page
Your agent's main page got a fresh layout that puts the most important info front and center — status, uptime, and quick actions. Configuration options now live on a dedicated Settings page so your overview stays clean.
Webhook settings show real use cases
The webhook panel now shows practical examples for connecting GitHub, Stripe, Zapier, and other services to your agent — with ready-to-copy curl commands and step-by-step instructions instead of raw technical details.
Schedule recurring tasks for your agent
The new Scheduled page lets you set up cron jobs that run your agent on a schedule. Pick from common presets — every hour, daily, weekly — or write your own cron expression. Edit, pause, or trigger any job manually with one click.
Admin dashboard for user management
Platform admins can now view all users, check gateway status, and manage accounts from a dedicated admin panel — no more kubectl commands to see who's running what.
Browse and install skills from the marketplace
The new Skills page lets you search the community skill library, see what's trending, and install plugins with one click — no terminal needed. Each skill shows its description and version so you know exactly what you're adding to your agent.
Smart onboarding checklist that adapts to you
The Getting Started section now tracks your actual progress and shows you exactly what to do next. Complete a step and it checks off automatically — connect your API key, link a messaging app, send your first message. No more guessing what's left.
Guided channel setup for WhatsApp, Telegram, Discord, and Slack
Connecting a messaging app to your agent is now a step-by-step walkthrough. Pick your app, follow the prompts, scan a QR code or paste your token, and you're live — the whole flow happens inside your dashboard without touching a terminal.
Live channel status and API key validation
Your agent page now shows which messaging channels are connected, disconnected, or errored — at a glance with color-coded status dots. Plus, the API key form now has a "Test Key" button that validates your credentials against each provider before saving, so you never wonder if a key is working.
Browser extension connects automatically
Your agent's browser now comes with the Browser Relay extension pre-installed, pre-configured, and pinned to the toolbar. No more manual setup — open the Browser tab and your agent can browse the web immediately.
Fixed connection lockouts under heavy use
A WebSocket connection counter leak was causing rate-limit lockouts (429 errors) during busy sessions. Connections are now tracked correctly so your agent stays responsive no matter how many tabs or sessions you have open.
New agents start with a free OpenRouter trial key
Every new agent now comes pre-loaded with an OpenRouter trial API key so you can start chatting immediately — no need to bring your own key on day one. The full OpenClaw onboarding walks you through upgrading to your own key when you're ready.
Approve or reject your agent's actions before they happen
When your agent wants to post on social media, send an email, process a payment, or deploy code, it pauses and asks for your sign-off first. The Pending tab shows exactly what your agent wants to do with full context — approve with one click or reject with a reason. You stay in control without micromanaging.
Activity feed now tracks every event type with detail panels
Click any event in your activity feed to expand a detail panel showing the full JSON payload, event type, category, token usage, and cost. Category icons and color-coded status badges make it easy to scan at a glance — lifecycle events in green, errors in red, tools in amber, social in teal.
Full product documentation at /docs/
BrainRoad now has 39 searchable documentation pages covering everything from first login to OpenClaw CLI commands. Six new pages go deep on the OpenClaw runtime — what it is, how to use the setup wizard, chat slash commands, console commands, channel connections, and configuration. Every doc page has sidebar navigation, breadcrumbs, and prev/next links so you can read straight through or jump to exactly what you need.
Daily briefings tell you what your agent did
The Activity page now has a Summary tab that generates a daily briefing — a plain-English recap of conversations handled, tools used, costs incurred, and errors encountered in the last 24 hours. Check in once a day and know exactly what happened while you were away.
See what your agent is doing — right now
Your agent page now shows a live activity pulse with real-time events, cost tracking, and category icons as your agent works. A full Activity page lets you filter by category, date range, and status — with a live WebSocket stream so events appear the instant they happen. No more guessing what happened overnight.
Every agent gets a public webhook URL
Your agent now has a dedicated webhook endpoint that any external service can POST to. Point GitHub, Stripe, Zapier, or any webhook-capable service at your agent's URL and it reacts instantly — no tunnels, no ngrok, no config. The URL and auth token are right on your agent page with copy buttons and curl examples to get started in seconds.
Update your agent on your schedule
When a new version of OpenClaw is available, you'll see an amber banner on your agent page showing exactly which version you're on and what's new. Click "Update Now" and your agent restarts with the latest version in about 30 seconds — no downtime surprises, no forced restarts.
Faster, more reliable agent sessions
Your agent's dashboard now connects directly to your gateway instead of routing through our API server. Sessions feel snappier, and platform deploys no longer interrupt what you're doing.
Try BrainRoad free for 30 days
Sign up and launch your agent instantly — no credit card required. You get the full platform for a month to connect your apps, test your agent, and see if it fits. After 30 days, pick a plan or your agent pauses until you're ready.
Two plans: Starter (free) and Pro ($29/mo)
Every new account gets the Starter plan — a full AI agent, all features, free for 30 days. After that, Pro is $29/month to keep your agent running. No hidden fees, no per-message charges.
Your agent can now install its own skills
Ask your agent "find me a skill for weather" or "I want to browse the web" through any connected chat app, and it will search the community skill library, show you options, and install your pick — no terminal needed. Every new agent ships with this built in.
Console quick actions that actually work
Redesigned the Console with compact action buttons that run real commands for you. Browse Skills now shows the most popular community skills sorted by downloads with clear instructions on how to install. Connect App walks you through linking WhatsApp, Telegram, Discord, Slack, or Signal step by step.
Two new guides for Console beginners
Never used a terminal before? We wrote two guides that explain the Console using concepts you already know — like pressing Next in a setup wizard. One covers what each button does, the other teaches you the five keyboard shortcuts that handle 90% of everything.
Signal messaging support
Your agent can now talk to you on Signal, alongside WhatsApp, Telegram, Discord, Slack, and iMessage. All six channels work out of the box — just scan a QR code or enter your credentials and you're connected.
All messaging channels enabled by default
New agents ship with WhatsApp, Telegram, Discord, Slack, Signal, and iMessage plugins pre-enabled. No config files to edit — just pick an app and connect.
Console stays stable during reconnects
The quick action buttons no longer disappear when your connection briefly drops. They dim while reconnecting and light back up when the link is restored — so you always know where you are.
Browser tab with live agent view
Your agent's browser is now on its own dedicated tab. Watch it browse the web, fill out forms, and interact with sites in real time — like looking over a coworker's shoulder. Includes a quick explainer so you know exactly what you're seeing.
Auto-approve device pairing
When your agent links a new messaging app, it no longer gets stuck waiting for manual approval inside the container. Device pairing and headless execution are now handled automatically so setup flows finish without intervention.
Quieter logs, faster dashboard
Reduced WebSocket proxy noise in the API logs and stopped unnecessary key-fetching calls when your agent is stopped. The dashboard loads faster and the logs are actually useful now.
Agent lifecycle you can trust
Start, stop, and restart your agent with confidence. We fixed several edge cases where the dashboard could show the wrong status during transitions. The boot and shutdown sequences now animate smoothly and reflect what's actually happening.
We ship every week.
Your agent gets better whether you're watching or not. Start free and see for yourself.
Get Started Free