July 2026

Everything that changed in NoClick in July 2026.

Improvements and fixes3
  • Running out of NoClick credits no longer points you at your model provider. The message now says your NoClick balance is the one that ran out, and its Add credits button opens the credits popup right where the error is shown instead of loading a new page. Buttons that link to a provider's billing page are now limited to agent nodes, so another service whose own error text happens to mention credits no longer offers one.

  • A CLI coding agent's sandbox now stops once your credits are used up, rather than running on past a zero balance. You get the credit balance email naming the workflow it was working on. Starting a turn on a CLI agent now needs at least 1 credit, so a turn that does start has room to finish.

  • The Credentials tab no longer turns red on a node wired into an agent as a tool provider when none of the operations you allowed needs a credential.

Improvements and fixes1
  • Changing an agent's model no longer empties the chat. The thread stays put until you send, and if the new model runs on a different harness or a different provider, the next message starts a fresh conversation that carries the earlier turns as context. Picking another model on the same provider continues the same thread and runs the model the picker shows.

Improvements and fixes1
  • The first click on Add trigger or Add tool under an agent's message now opens the picker. It used to take two clicks.

Improvements and fixes1
  • Agents that chat on WhatsApp, Discord, Microsoft Teams, and other channels now reply to the right person and keep a separate memory per conversation. Replies that used to go nowhere now deliver.

Improvements and fixes2
  • The list of steps an agent took in a chat reply stays in place after you refresh the page or switch tabs.

  • An agent no longer runs when a step feeding into it has failed, so it stops instead of acting on missing or error data.

Improvements and fixes2
  • Public shared agent chats no longer show a spurious "still working on the previous message" error when you send a quick follow-up, and long conversations keep the message box in view instead of scrolling it away.

  • Agent replies read more cleanly: the run-results popup formats markdown instead of showing raw text, and messages that mention dollar amounts no longer render as math.