Viewpoint
n8n AI agents, and the part nobody demos.
An n8n AI agent takes an afternoon to demo and a quarter to trust. The gap is tool contracts, memory, evals, and blast-radius limits.
Notes from agents running inside legal, finance, SMB operations, and enterprise IT.
02 / The build
What an n8n AI agent is made of.
Tool contracts
Every tool gets a typed input, a narrow permission, and a description the model can act on without guessing. Vague tool descriptions cause more agent failures than weak models do.
Memory that earns its keep
Short-term state in Redis or Postgres, retrieval over pgvector for the documents that matter. Persist decisions and identifiers, not raw transcripts.
Bounded loops
Max iterations, max tool calls, and a hard timeout. An unbounded agent is a billing incident waiting for a Friday.
Approval gates
Human-in-the-loop steps on anything that moves money, contacts a customer, or writes to a system of record.
Retries and idempotency
Exponential backoff, dead-letter queues on every webhook, idempotency keys on every external write, so a retry never double-charges anyone.
Traces and evals
Full execution history plus a fixed eval set that runs on every change. If a prompt edit regresses the case set, it does not ship.
03 / Agent or workflow
Most n8n agents should have been workflows.
If the steps are the same every time, a deterministic workflow wins on cost, latency, and debuggability. Reach for an agent when the input is unstructured, the path genuinely varies per case, and a human would need judgment to route it. That test kills about half the agent ideas I get asked about, which is the point.
04 / Failure modes
What actually breaks in production.
Tool descriptions the model misreads. Retrieval that returns plausible but wrong context. Loops that retry a failing tool until the budget is gone. Prompt injection arriving through a document the agent was asked to summarize. Silent success, where the agent reports done and nothing changed downstream. Each of these has a boring, specific fix, and none of them show up in a demo.
05 / Cost control
Spend you can attribute to a workflow.
Prompt caching on stable system context, smaller models for classification and routing, the strong model only where reasoning is load-bearing, and a per-workflow cost dashboard. Iteration caps do more for the bill than any model swap.
06 / Proof
Agents that other people operate.
I have shipped n8n agents inside legal, finance, SMB operations, and enterprise IT, then handed them over with a runbook, recorded walkthroughs, and a week of pairing. If nobody can run it without me, it is not finished. Completed 9 Anthropic Academy courses. Lovable Champion and community leader.
FAQ
Common questions.
- What is an n8n AI agent, exactly?
- A workflow where a model decides which tool to call next, instead of you hard-wiring every branch. n8n holds the tools, the retries, the state, and the approval gates. The model holds the judgment. Everything else is plumbing you still have to get right.
- When is an n8n AI agent the right shape?
- When the input is messy, the steps vary per case, and a deterministic workflow would need dozens of branches. If the path is fixed, skip the agent. A plain n8n workflow is cheaper, faster, and far easier to debug.
- Which nodes do you actually use?
- AI Agent and Tools Agent nodes for orchestration, Anthropic Claude for reasoning and structured extraction, MCP client nodes for tool exposure, Postgres with pgvector for retrieval, and Redis or Postgres for memory. Webhooks in, queues behind them.
- How do you stop an agent from doing something stupid?
- Blast-radius limits. Tools are scoped to the narrowest permission that works, writes go through idempotent endpoints, and anything touching money, customers, or compliance hits a human approval step before it commits.
- How do you know it still works next month?
- Evals on a fixed case set, run on every change, with a pass threshold that blocks a deploy. Traces on every execution so you can read what the agent chose and why. Without both, you are guessing.
- What does an n8n AI agent cost to run?
- Token spend scales with tool-call loops, not with request count, so a badly bounded agent can cost ten times a well bounded one. I cap iterations, cache prompts, and put a cost dashboard per workflow so spend is attributable.
- Self-hosted or n8n Cloud for agents?
- Self-hosted when data sensitivity, queue mode, or custom nodes demand it. n8n Cloud when speed matters more than control. Agents run fine on both, but queue mode matters once concurrency is real.
- Where does the paid work live?
- On TechTide AI. This page is the field notes. If the agent is stuck between demo and dependable, the Systems Decision Audit is the way in.
If the agent works in the demo and nowhere else.
That is a decision problem, not a code-volume problem. The Systems Decision Audit at TechTide AI returns ship, fix, redesign, or stop, with the reasoning written down. When the data underneath the agent lives in Airtable bases, the Airtable Rescue Decision Sprint is the right first move instead.