People ask me “why would I use Eliza over n8n?”
You wouldn’t. You’d use them together.
Here’s a cool thing: LLMs generate n8n graphs really well. The graphs are just JSON.
LLMs use the n8n graphs as a short hand for data flow and extensive API access. n8n cuts a lot of boilerplate and reduces the need for another whole ecosystem of connectors to every API.
The agent requests API keys and config as needed for the admin, which can be a human or another agent tasked with the job.
All of this goes for MCP as well of course, and the agent can drop down to the level of writing pure code, using tools like langchain and instructor.
None of this stuff is in competition, all of it is highly synergistic. I think n8n is still too big brain for normies to use but just right for a lot of companies that just need a workflow, or as a go to for an agent to generate a solution to its problem as needed.