n8n Launches AI Agent in Early Preview to Run Alongside Workflows.

n8n has released an AI agent that runs natively alongside its workflow automation platform, now available in Early Preview. The feature marks a shift from purely visual workflow building to autonomous AI-driven automation within the same environment.

The agent is designed to operate as a "fully capable" companion to existing n8n workflows, according to Leonardo Grigoryan, product lead at n8n. The company has published documentation and self-hosted setup instructions for teams running their own instances.

n8n's implementation differs from standalone agent platforms. Rather than forcing users to choose between visual workflows and agentic automation, the agent lives within the same interface as traditional workflows. This means teams can decide when to use structured logic and when to delegate tasks to an AI agent that can reason through variable scenarios.

The agent can interact with the same APIs and services that n8n workflows already connect to, using the platform's library of over 400 integrations. For technical teams already running n8n workflows, this removes the need to stand up separate agent infrastructure or learn a new toolset.

Self-hosted deployment is supported from launch, consistent with n8n's open-source positioning. Organizations running n8n on their own infrastructure can deploy the agent without sending data to external services, a constraint that matters for compliance-sensitive workflows.

n8n has shipped several AI-focused templates in recent months, including tools for invoice processing and fashion try-on workflows. Those templates used AI within predefined workflows. The new agent feature, by contrast, allows AI to make decisions about workflow execution itself.

This positions n8n closer to platforms like Zapier and Make, both of which have added AI-driven automation layers. The difference is architectural: n8n's agent runs on the same self-hostable infrastructure as its workflows, while competitors typically route agent requests through their cloud services.

The Early Preview label signals that the feature is not production-ready for all use cases. n8n has not published benchmarks on agent reliability, cost per execution, or failure modes. Teams testing the agent should expect changes as the feature matures.

For teams already using n8n, the agent expands what automation can handle without custom code. Scenarios that required brittle conditional logic can now be delegated to an agent that interprets context. For example, triaging support tickets, summarizing meeting notes, or routing data based on unstructured input.

For teams evaluating automation platforms, n8n's agent feature adds a third option between low-code tools like Bubble and full-code environments like Replit. Workflows handle predictable tasks; agents handle ambiguity.

The feature is available now in n8n Cloud and for self-hosted deployments. Documentation covers setup, agent configuration, and integration points. As with all Early Preview features, teams should test thoroughly before deploying to production workflows.