Self-healing runs
When a run fails, Decisional triages the failure, patches code-level bugs, retries transient errors, and only asks you when a human decision is needed.
Agentic workflows. Zero API pricing. Powered by your Codex or Claude Code subscription.

Build automation agents in Decisional which each have their own workflow. When it reaches an AI Agent node, your Mac runs Claude Code or Codex locally.
When a run fails, Decisional triages the failure, patches code-level bugs, retries transient errors, and only asks you when a human decision is needed.
Switch between Chat, Operator, and Build without losing the thread, so one agent can talk, operate, and change its own workflow.
Watch the workflow run step by step with logs, output previews, and run context where the work is happening.
Move between Chat, Operator, and Build from one picker while the agent keeps the same thread.
Pause execution, review the context, and approve the next action before the agent continues.
Tool calls, results, and errors stream directly onto the workflow canvas as the run progresses.
A tiny arcade game keeps the workspace alive while your agent generates and validates the workflow.