August 2026

The popup that opens when a run finishes now walks through the run in reading order instead of listing node outputs one at a time. What came in shows the event that started the run in its own app's frame, so a WhatsApp message looks like a WhatsApp message and an invoice looks like an invoice. What the agent did lists the agent's tool calls with the time each one took and the detail that matters on the row, such as the channel a message went to, and any row opens to show what it was Called with and what it Returned. What went out frames each message the agent sent the way its destination shows it. An agent that worked and chose to send nothing says so in its own words, and a failed run shows the error with a way to fix it instead of a verdict.

The run switcher at the top of the popup reads Latest run and opens your run history: each row has its status, what fired the run, how long it took, how many nodes ran, and the error on runs that failed. It stays reachable while a run's results are still loading and on a workflow that has never run, so history no longer dead-ends.

Type @ in an agent chat and a list of the conversation's workspace files appears. Pick one and its full path is inserted into your message, so you can point the agent at a file without typing the path out. Arrow keys move through the list, Enter or Tab accepts the highlighted file, and Esc closes it.

You can also delete files now. Each row in the Files menu has a trash button that asks you to confirm on the row itself before the file is removed from the workspace.

Every agent on a workflow now has a Linux sandbox with a shell, whether or not you have wired a filesystem node to it. The sandbox starts the first time the agent actually runs a command, so an agent that only ever answers in chat never starts one. In a chat, the agent's /workspace folder is kept per conversation and survives across runs.

A test run answers shell commands with made-up output like every other tool call, so a rehearsal never starts a real sandbox.

The AI builder can look things up for itself while it builds too. Paste an API URL or a sample payload and it can fetch the URL and inspect the real response instead of wiring a node blind. It has no access to your connected accounts while it does this.

You can now upload files into a workspace yourself, so the agent finds them there on its next turn. In an agent chat, use the upload button in the Files menu header, and the files land in that conversation's workspace. In a filesystem node's config panel, use Upload in the file browser, and they land in that node's volume.

Every page under noclick.com/agents now has a Setup section under the hero that walks the whole path before you commit to it: choose whether the agent runs On your computer or In the cloud, pick what it can do in each app, then sign in to each app once.

The last step follows your choice. On your computer ends on the exact command that connects your coding agent to NoClick, shown with a placeholder link because setup mints the real one. In the cloud ends on signing in to the coding agent NoClick runs for you.

Click Test in an agent's chat header to run it against a staged event. Your real agent runs on your real model and prompt, but every tool call it makes is answered with made-up data, so nothing reaches your accounts and nothing is sent. The screen says so up front, and every answered call carries a simulated pill.

Gmail, Slack, WhatsApp, and Telegram triggers ship written situations to test against, and every other trigger gets an editable sample. You can rewrite the staged message in place, keep several named runs per agent, and share a finished run as a read only page anyone can open.

Template pages are now built around the agent. Each one previews the real canvas, lists the services setup will ask you to connect and the questions it will ask, and shows the staged conversations the template ships so you can see the agent work before you take a copy. Click Use this agent and your copy opens straight into full screen setup. The gallery is now Agent Templates.

An agent node can hold connected keys for more than one provider, and switching the Model could leave an earlier key attached. That key was used anyway, so the run failed on a rejected or missing API key even when the right key was connected.

NoClick now runs the agent on the credential belonging to the provider of the model you picked. A key for any other provider is ignored, and the run behaves as if nothing were attached: models NoClick serves on its own keys keep running on credits, and models that need your own key stop with a missing credential you can fix in the node's Credentials tab.

The agent chat in a workflow's Interface tab now lays itself out from the width it actually has, rather than the width of the window. When something else takes up the screen, like the builder drawer or a split view, the chat keeps its full width and Settings opens as a sheet over it instead of splitting the row.

• History, Files, Share, and Settings in the chat header collapse to icons when the header is too narrow for their labels.
• The History, Files, and model pickers open as full-width sheets on phones instead of cramped dropdowns.
• Tablets in portrait use the single-column tab layout, so the navbar and canvas are no longer squeezed into a sliver.

The Add trigger and Add tool palettes on an agent node now search each service's description, its aliases, and the actions it can perform, not just its name. Searching for "schedule", "timer", or "every day" finds the Schedule trigger, and "create issue" surfaces Linear, Jira, and GitHub.

Services also carry the same name here as on the canvas. Schedule used to be listed as "Trigger Cron" and X as "Twitter", and about a third of the catalog appeared under a name that did not match the one you see everywhere else. Once you pick a service, its list of triggers is searchable by keyword too.

A CLI coding agent's sandbox is billed for as long as it stays up, so a turn that started and never delivered a reply used to keep costing 0.05 credits per minute until the four-hour sandbox limit. Those turns now end: the sandbox shuts down as soon as the agent's process is gone, and a turn that produces no output at all for 30 minutes is treated as stalled and stopped.

• The chat reports the turn as lost and unlocks, so you can send the message again, instead of showing a reply that is still coming and never arrives.
• Files the agent saved to its workspace survive the stop, as they do for any other sandbox shutdown.
• A turn that is genuinely quiet for 30 minutes, for example one long build inside a single command that prints nothing, is stopped as well and has to be sent again.

A photo, voice note, or document sent to a WhatsApp-triggered agent now arrives as a link the agent can download, along with the file type and name. The message caption comes through as the text. Media used to reach the agent as a placeholder with nothing behind it.

• Files up to 25 MB are saved to your workflow's resources at the moment the message arrives, so the link keeps working for the whole run.
• Larger files, and files the provider no longer holds, are reported to the agent as media it could not retrieve, so it can ask the sender to resend instead of guessing.
• Nodes wired downstream of the WhatsApp trigger read the same link from the trigger output.