Skip to main content

Creatio AI Studio updates: October 8, 2026

With this update, you can talk to AI Twin instead of typing. Record a message, and AI Twin turns it into text you can edit before you send it. AI Twin also saves your agent changes as a draft and deploys them only when you ask. You can dictate in the test chat of an agent too. Learn more about all the new capabilities below.

AI Twin​

Dictation in the AI Twin chat. You can speak your message to AI Twin instead of typing it. Click Voice message in the message box and record your message. Then click Stop and insert text. The text appears in the message box, and you can edit it before you send it. To start over, click Discard recording. Nothing is sent until you send it yourself.

Agent changes saved as a draft by default. You decide when an agent goes live. When you ask AI Twin to create, change, or fix an agent, AI Twin saves the changes as a draft. If you did not ask to deploy, it offers to deploy them as the next step. AI Twin publishes and deploys the agent only when you ask for it, for example, when you answer "yes" to its offer. It deploys to your current environment, or to the environment you name.

Skills​

Size limit for skill instructions. To keep your skills efficient, Creatio AI Studio limits the size of skill instructions. The SKILL.md file can have up to 21,856 characters. When your skill exceeds this limit, move the details into files in the references/ folder.

Skills in realtime voice conversations. Agents can use their attached skills in realtime voice conversations.

Agent evaluation​

Dictation in the test chat. You can record a message in the test chat of an agent instead of typing it.

More time for multi-turn eval cases. An eval case gets 120 seconds for each user turn it sends. A case with three turns can run for up to 360 seconds.

Channels and integrations​

Tool status per environment for MCP integrations. An MCP integration can have a different URL in each environment. The Tools tab shows which tools are ready in each environment. A tool that was not discovered in an environment is marked "Not yet discovered — run Discover Tools for this environment." A tool that cannot be reached is marked "Not reachable in this environment."

Observability​

Clearer run and session metrics. The Model calls tile shows the total number of model calls. This includes calls that the platform makes on its own, such as memory updates. When there are such calls, the line below the total splits it between the agent and the platform. The session duration tile is called Active duration.

Usage in AI credits. Run and session metrics show usage in AI credits. This includes the tiles, the model call details, and the list of sessions.

Usage and budgets​

Dictation recorded in AI credits. Speech-to-text for dictation in the test chat is recorded in usage, together with its cost in AI credits.