Changelog

Your assistant's email address used to take only your own mail. It now takes anyone's, so people and other agents can write to it. Mail from someone else never lands in your own thread with it: each sender's mail gets a conversation of its own.

Every account has its own invite link, under your balance in the sidebar. When someone you invite starts using NoClick, $2 is added to your balance.

Settings → Usage lists what your account was actually charged, in dollars, with a plain name for each charge, and drills down to what used it: which agent, app, number or automation.

From a domain's page in Resources, pick one of your websites to serve there. Each name says where it stands: registering, DNS, certificate, then live. A purchase that fails frees the names it was going to serve.

Every connected account now carries its own approval rules, and they apply to everything that uses it: your agents, your coordinator, and any MCP client. Open Settings → Credentials and click the shield on a connection to reach its Tool permissions page.

A locked action runs nothing until you say so. The call waits, the exact arguments it would send are shown to you, and you approve or decline it.

• The Use without approval row sets everything at once. Read-only lets reading actions run and locks the rest, All lets every action run, and None locks everything, including actions the service adds later. Search the list to lock one action on its own.
• Requests arrive under Needs you on the Dashboard, where you can read the arguments and decide without leaving the page.
• Approving admits one call with exactly those arguments, and it has 30 minutes to run. Different arguments, a second attempt, or a change to the rules each need a fresh approval. An untouched request expires after 24 hours.
• Agents and MCP clients can add restrictions but never remove one. Unlocking happens only on the permissions page, signed in as the connection's owner.
• Operations you allowlist on a tool provider node now carry a lock in the node settings, so you can see and change what needs approval without leaving the workflow.

Every account now has a coordinator: one assistant that sees the whole account rather than a single workflow. It reads your workflows and runs, hands builds and fixes to the AI builder, passes work to agents you already have, and connects accounts for you.

• It watches failed runs. When a workflow fails, the coordinator looks into it: it asks the builder to diagnose and fix, pauses a workflow that is clearly finished with, or tells you when only you can fix it. A run that failed for want of credits never wakes it, because the credits alert already covers that.
• Memories is a new card on the Dashboard. It holds what the coordinator has learned about you and your work, as entries you can read, edit, write yourself, and delete. Each one is filed as About you, Preferences, Project context, or Reference.
• Preferences live there too, rather than in a settings page. Tell the coordinator to stop triaging a workflow, or which channel to reach you on, and it keeps that as a memory.
• It reaches you where you last wrote to it, on WhatsApp or by email. It has its own inbound address on noclick.app, so you can mail it and it answers on the same thread.
• It makes images and video on request. Video generation is available on the Plus and Pro plans.
• Your agents can message it as well, so an agent that hits something it cannot fix has somewhere to send it.

The sign-in page and the in-app sign-in panel now offer Continue with phone. Type your number, tap Text me a code, and enter the code to sign in. A number NoClick has not seen before makes an account, so the same form covers signing up. The number field has a searchable country picker, adds the calling code to digits typed without one, and formats the number as you type.

The number you message NoClick from on WhatsApp signs you in to the same account, so a chat you started on your phone and the workflows you open in the browser are one account rather than two.

• Phone is now a section in Settings for every account. Add a number, prove it with the code we text you, and WhatsApp messages and calls from it reach your account. Codes expire after ten minutes. Each row says whether the number was Verified with a code or Linked on WhatsApp.
• An account can link up to five numbers, and a number belongs to one account at a time. Unlink frees a number for another account to claim, and a number that is the only way into an account cannot be unlinked until the account has an email address.
• A chat that started on WhatsApp with no account behind it can be connected to an account you already have, from inside the chat. Ask to connect the email you sign in with, and NoClick emails that address a six digit code for you to send back. The code expires after fifteen minutes and is spent after five wrong tries, and nothing moves until it matches, so naming an address you do not hold reaches nothing. On a match, your workflows, connected accounts, and the chat itself move to the account with that email. Where no account has that address, the phone account takes the email so you can sign in with it too.

Pasting an API key used to save whatever you typed. A key with a stray line break, a web address pasted into a token field, or a key the service does not accept all saved looking connected, and you found out at the first run. NoClick now tidies the value and asks the provider about it before storing anything.

When the provider answers, the form shows what it said about your account rather than a tick: your own boards, sites, or workbooks, or Connected as and the account name. When the provider refuses, nothing is saved, and you get its own words along with what that usually means, such as an address serving a sign-in page instead of an API, or a key that reads as truncated or revoked. When NoClick cannot reach the provider at all, the credential saves and says Saved, not verified rather than claiming to be working.

• Problems are reported under the field they belong to. A web address pasted into a key field, placeholder text left in, and a value the service's own rules reject each name the field, and clear as you edit it.
• Credential forms now show the setup instructions and the per-field help for the service. Both were already written for each service and neither was displayed.
• A credential that proved itself when you connected it is remembered as Verified, with the account it connected as, so a node's Credentials tab says more than that something is attached.
• Tableau accepts your site address as it appears in your browser. Paste the whole URL into Server URL and NoClick fills in Site Content URL from it. The Cloud Manager console address is refused when you save, because it never serves the API, and the message names what to use instead.
• The same check runs wherever a key is entered: on a node, in Settings → Credentials, during Setup, and on a link someone sent you to connect an account for their workflow.

The Setup tab now ends with a check on the workflow itself. It reports one of six states: setup unfinished, configured but unverified, registered and waiting for a real input, a reporting deadline armed, a real input received, or nothing automatic wired at all. When an input has arrived it names the node, when it arrived, and what the run did with it, and a Refresh button rechecks without reloading the page. It also says outright that a Test Run is simulated and that a send the provider accepted is not proof it was delivered.

• A trigger registers only once the path it starts can run. A trigger whose workflow is still missing an account, an agent's instructions, or a required field no longer shows as live while nothing could have run. Fix what is missing and the trigger registers itself the next time the workflow saves.
• A WhatsApp trigger that could not register says why, for example that the saved phone link is disconnected and needs reconnecting, and the message clears once a later attempt succeeds.
• A trigger run that carried no event stops there. A push trigger you run by hand reports that it is waiting for a message and no longer hands an agent an empty turn to spend credits on. Branches that do not depend on that trigger still run.

Message is where you write an agent's standing instructions, and it is still required when nothing feeds the agent. When a trigger is wired into the agent, the event that arrives is the turn, so Message can be left empty and the node is no longer marked incomplete for it. Running such an agent by hand, with no event and no standing instructions, fails with an error saying so rather than running on nothing.

A required field holding a reference like {{ $('trigger').chat_id }} used to be passed on as an empty value when the referenced output had no such key. The step reported success and nothing arrived. That field now fails the step, and the error names the field, the reference, and the keys the referenced node's output actually has, so you can fix it without opening the run.

• The error reads Required field 'to' is empty, followed by what went wrong: the referenced node has no key by that name, and the list of keys it does have. A key that exists but holds null is reported as null, which reads differently and usually means a different fix.
• A reference read during a manual run of a trigger that waits for events says there is no live event in this run, and that firing the workflow for real is how to test that path.
• Optional fields are unchanged. A reference that resolves to nothing there leaves the field at its default, as before.
• Sending a WhatsApp message to an empty recipient stops with a message naming the chat id to reference instead, rather than being accepted and never delivered.

Slack sends channel events only to apps that are members of the channel. A trigger on a channel NoClick was never invited to used to read Active and never fire. Registration now checks, and the trigger's Status field carries the verdict. When the app is not in the channel, the status turns amber and says so, with a Join #channel button next to it on a public channel. A private channel names the invite to run instead, since only a person can add an app to one.

• The node is marked incomplete on the canvas, counted by the incomplete pill at the bottom left, and named in the Setup tab. It never holds up a run you start yourself, because a manual run does not wait for Slack to deliver anything.
• Your Dashboard lists it under Needs you as not receiving events, with that same fix as the button on the row and Open trigger next to it. Open trigger now opens the workflow with the trigger selected and its settings showing.
• Ask the AI builder why a workflow is not running and it reads the same verdict, so it answers that the trigger is registered but not receiving events rather than calling it active.
• A check that cannot reach Slack changes nothing. A trigger is never called broken on a failed check.

A scheduled workflow that kept failing the same way used to mint another failed run on every tick, for as long as the schedule ran. NoClick now pauses it: when the last five runs from that schedule have all failed with an identical error, spread over at least 30 minutes, the trigger node is switched off and the schedule stops.

• You get one email naming the workflow, the paused trigger, the error, and when it was paused. Paused schedules is its own toggle in Settings → Notifications, and you get at most one of these per trigger per day.
• The paused trigger appears on your Dashboard under Needs you and on the Triggers card, named and marked as paused with the error next to it. A trigger you switched off yourself stays hidden, as before.
• Fix what the error names, then enable the trigger node again to resume the schedule. If the same failure comes back, the schedule pauses again on the first repeat.
• Running out of credits never pauses a schedule. Topping up resumes the runs on its own.
• A short provider outage does not reach the bar: the failures have to be identical and sustained for half an hour.

The connect status for a workflow that hosts an MCP server now answers two questions separately instead of showing one spinner for both. A Link ready line names how many tools the link serves right now, with a few of them by name, read straight from your workflow with no client involved. Under it, the client line reads Waiting for your agent to connect until a client calls in, then Connected with the client's name, how many tools it was served, and which ones it can call.

• The waiting line names what makes that client connect. Most clients only write the config when you run their add command and open the connection when a session starts, so the usual fix is to start or restart the client rather than to wait.
• After 45 seconds the panel offers that client's own check command, such as claude mcp list or opencode mcp list. Those commands list the servers and connect at the same time, so running one turns the status green.
• The status also sits on the node. The Connect externally header of an MCP Server node reads Connected, with the client and the tool count, for anyone who opens the workflow later, so you do not have to run the Setup finale again to check.
• When NoClick cannot read handshakes for a moment, the panel says it cannot check rather than reporting that nothing has connected. The link keeps serving either way.
• The client setup guides are re-verified. Claude's custom connectors live under Customize → Connectors, Zed connects to a remote URL without a bridge, and VS Code adds the server with code --add-mcp.

The Feed tab in the top navigation becomes Dashboard: one page for what is happening across your workspace, with a card for each area and a full page behind every card. Needs you replaces the approvals queue and now also carries questions the AI builder parked, edits an agent proposed to its own workflow, accounts someone still has to connect, connections that stopped working, and triggers that failed to register. You answer each one where it sits, with the same fields and pickers the workflow itself would show you.

The other cards are Runs, Agents, Upcoming, Files, Credentials, Triggers, Credits, and Notifications. Press G then D to open the Dashboard, or G then I to land on Needs you.

• Runs charts the last 14 days and lists your workflows by run count, with failures called out. Click a row to open that workflow, or a single run to see its results.
• Files gathers workflow uploads, agent workspaces, and the builder's scratch folder in one place, previews images, video, audio, and datasets without leaving the page, and lets you upload or delete wherever you have write access.
• Agents shows sandboxes that are up now and each agent's latest turns, with the message that came in, the messages the agent sent, and every tool call it made.
• The section you drilled into is remembered while you visit other tabs. The back button, browser Back, or Esc leaves it.