Runchat Updates: May 8, 2026

🪢 AGENT SWARMS, ONBOARDING WIZARD, RUN PERMISSIONS

• When you use the canvas agent for the first time you now get walked through how to configure the agent for your use case. This helps you make good tradeoffs between cost, speed and quality. You can always re-run the wizard by clicking the button at the top of the chat settings.
• One of the things the onboarding suggests is to give the agent Run permissions. This is very useful if you're creating scripts where running the node costs no credits. The agent can write scripts, run them and then check the output to verify that everything works. No more copy pasting errors or asking for bugs to be fixed.
• The agent can now run python scripts (this would error out previously)
• We now support running multiple agents on the canvas at once. You can create a new agent by clicking the "New" button in the chat input.
• Multiple agents let you work on multiple scripts at once, kick of multiple research tasks and more.

🍎 APPS WITH STATE (Alpha)

• You can now use the "App" output mode with the code node to create apps that have a front and backend with their own apis.
• Apps can be written with multiple files, making it easier for LLMs to manage more complex code
• This is useful for building anything that needs to remember things over time for your user (e.g. todo lists and the like)
• Apps are in alpha as we work out use cases and costs, and are limited to hobby and pro subscriptions.

📄 BETTER PDF SUPPORT

• You can now upload pdfs more easily
• You can read these with any agent model in chat or with the agent node

🪲 OTHER IMPROVEMENTS

• We now show the rhino toolbar in the plugin even when the window is very narrow
• Added timer to agent tool calls and thinking text
• We now support much larger files on the canvas (<30mb)

Whats coming next? More support for creating apps with the agent and managing automated tasks.