Gumloop

Category: Automation/Workflow

Agent platform for marketing ops. The free tier is gone.

Gumloop has completed its shift from a visual automation canvas to an AI agent platform; the canvas is now labelled legacy. You describe what you want and build agents that run on a schedule or on demand, with Skills that sync to GitHub, a Company Brain for organisational context, evaluations for testing agent output, an agent browser, a CLI and SDKs. Typical marketing uses are lead enrichment, content repurposing, competitor monitoring and report generation. It both ships and consumes MCP: there is a remote server exposing around eighty-seven tools covering agents, sessions, skills and artifacts, and you can also connect external MCP servers as agent capabilities, with hosted and proxied options on Enterprise. Two commercial changes matter: the free tier has been removed, so the floor is now a paid plan, and there is an orchestration fee charged on top of model and tool costs.

Our verdict

One of the better agent platforms for non-engineers, and the bidirectional MCP support is genuinely strong: eighty-seven tools exposed, external servers connectable as agent capabilities, and a documented pattern where Gumloop holds the credentials so they never reach your client. For marketing operations work it is more approachable than n8n and more capable than Zapier. Three things to factor in. The free tier has gone, so there is no way to try it without paying. Credits do not roll over, there is an orchestration fee on top of underlying model costs, and the documentation openly admits the same agent might cost two credits or two hundred depending on the task, which makes fixed-fee client delivery genuinely risky. And the one that matters most for agencies: on self-serve plans including the paid entry tier, your content may be used to train Gumloop's models, with only Enterprise excluded. If you are handling client data, that is a blocker rather than a footnote.

Best for: Marketing operations people who want to build working agents without writing code, and who have budget for a paid plan from the start.

Pros

  • Ships a remote MCP server with around eighty-seven tools, and consumes external ones too
  • Credentials can be held by Gumloop so they never reach your AI client
  • More approachable than n8n, more capable than Zapier, for marketing ops work
  • Skills sync to GitHub, with a CLI, SDKs and evaluations for testing agent output
  • Company Brain gives agents organisational context
  • Unlimited seats and unlimited agents on the paid plan

Cons

  • Self-serve plans, including the paid entry tier, may have your content used for training
  • Only Enterprise excludes training, which is a blocker for agencies handling client data
  • The free tier has been removed, so there is no way to trial it without paying
  • Credits do not roll over, and an orchestration fee sits on top of model costs
  • Their own docs admit the same agent can cost two credits or two hundred
  • US-only processing by default, with EU residency only via an Enterprise arrangement
  • Audit logs, RBAC and usage visibility are all Enterprise only

Pricing: From $37/mo, no free tier (Paid)

Rating: 4/5