Gumloop vs n8n: Credits, Executions and Hosting Compared

Gumloop and n8n solve the same problem with two different meters. Gumloop is a hosted platform where a team describes AI agents in plain language and pays for their work in credits: Pro is $37 a month for the whole organization's shared pool of 20,000 credits.^[1] n8n is workflow automation on a visual node canvas, billed per workflow execution, and it runs on n8n Cloud or on a server you manage yourself. Pick Gumloop when business teams want agents inside Slack or Microsoft Teams and nobody wants to run a server. Pick n8n when a technical team wants self-hosting, code steps and a count of whole runs, with model calls billed apart. The two meters do not convert into each other.

Credits and executions measure different things

Every Gumloop agent chat is priced mainly from four lines: model tokens at the provider's list price, at least 1 credit for each successful connector call plus the tool's own charge for data services such as Apollo or Firecrawl, 5 credits per session-minute of active compute, and an 8% orchestration fee on top.^[2] Brain searches, evaluations and voice add their own lines when used. Gumloop says the same agent might spend 2 credits on a quick question and 200 on a deep research task. Failed connector calls and time spent waiting for you cost nothing. The bill tracks effort. A longer conversation, a heavier model or more tool calls all raise it, and thinking tokens count too.

n8n counts something else. An execution is one complete workflow run from start to finish, whatever the number of steps, with no per-step or per-user charge, and users and workflows are unlimited on every plan. A turn with a standalone n8n agent counts as one execution, and calls to sub-agents or workflow tools add nothing. n8n Cloud Starter costs $20 a month billed annually or $24 a month billed monthly for 2,500 executions a month.^[3] n8n Cloud Pro costs $50 a month billed annually or $60 a month billed monthly for 10,000 executions a month.

Model calls sit outside that count. On n8n Cloud Starter and Pro, nodes can call AI models without your own API keys through Gateway credits, a prepaid balance billed per request on top of the plan (new Cloud users get a small free amount at sign-up); otherwise you pay your own model provider. Gumloop's pricing page prints only a monthly Pro price and shows no annual plan, so annual figures on the n8n side have no Gumloop counterpart to set against them.

A documented example shows how the credit side adds up. By our arithmetic, the docs example run of 459 credits (425 credits of work plus a 34-credit fee) comes to about $2.30, and the same run with your own model key drops to 93 credits, about $0.47 plus whatever your model provider bills for the tokens, because the fee then rises to 16% of the full list cost while tokens cost no credits.

That is one run, not a typical month: your own key saves credits on token-heavy runs, but on runs that are mostly tool calls or compute time it can add credits. An execution count ignores what happens inside the run, though its model calls are billed apart. Size it with a pilot. Count runs and model spend on one side and read the per-chat cost lines on the other before you compare monthly bills.

Running out works differently on each side. Gumloop's 20,000 credits reset monthly with no rollover outside Enterprise, overage at $0.005 a credit is off by default and capped on Pro, and agents and triggers stop when credits or the cap run out; our Gumloop pricing breakdown covers every billing rule. On n8n Cloud, runs over the monthly execution limit fail immediately, without being queued, until the count resets on the 1st of the month or you upgrade.

Hosted service or your own server

Gumloop is sold as a managed service, and users outside the US have their data transferred to and processed in the United States. On Enterprise, Gumloop can be deployed into the customer's own AWS, Azure or Google Cloud account, in a region of the customer's choice, and Gumloop deploys and operates that stack. There is no self-hosted Gumloop for individuals. The closest option is that Enterprise VPC deployment. Gumloop is not open source, and the open-weight models it offers are models, not Gumloop's own code.

n8n gives you more places to run. Its Cloud instances are hosted on Azure in the EU (Frankfurt), though AI features such as Gateway credits use sub-processors that process data in the US, and the Community Edition is free to self-host, which means you run and pay for the server yourself. n8n Business costs $800 a month billed annually or $960 a month billed monthly for 40,000 executions a month, and it is self-hosted only, not a Cloud upgrade. Bash scripts and custom nodes work only on self-hosted n8n. n8n says self-hosting keeps data in-house. n8n calls its license fair-code and source-available rather than open source.

Trials differ as well. Read the card terms. Gumloop has no free plan: every new account starts with a one-time 14-day Pro trial that needs a card and becomes a paid Pro subscription unless you cancel before it ends. n8n's Cloud trial runs 14 days with no card, while every Business price carries a 14-day trial that requires a payment method. Under Gumloop's Terms, paid subscription fees are non-refundable except where the law requires otherwise, and canceling keeps your agents and data in the account. n8n refunds a monthly Cloud subscription if you ask within 14 days of the billing date and cancel the account, while annual plans are non-refundable.

Agents you describe, or workflows you wire

A Gumloop agent is set up with instructions, a model, connectors, skills, Brain knowledge, subagents and triggers, either in the agent editor or by asking Gumball, the personal agent every account gets, which builds the agent after you approve a card. Every agent also has a built-in computer for Python and shell commands in an isolated cloud machine, plus a cloud browser that can sign in, fill forms and download files. No code is required. Approvals are set per connector: always allow, ask each time, ask only for writes or deletes, or never allow.

Gumloop's older step-by-step builder, Workflows (once called Flows), still appears on the pricing page marked Legacy; Gumloop's current docs and product are built around agents. Gumloop says agents can call workflows. On Pro, Gumloop allows 25 concurrent agent chats, and requests over that limit are rejected while triggers are skipped. Scheduled, app, webhook and AI-created triggers start agents without anyone typing. AI-created triggers poll every 5 minutes by default and bill every poll, so one left at that interval draws on the pool whether or not it fires.

n8n starts from the canvas. Its AI Agent node, MCP client, human approval for tool calls and hosted chat are on every plan. Agents are still labeled Preview. Existing workflows become tools an agent calls, and a workflow can call a published agent through a Message an Agent node. Code steps in JavaScript and Python work on every plan, though on n8n Cloud they cannot import outside libraries (only on self-hosted n8n). Starter allows 5 concurrent executions and a 5-minute maximum run, while Pro allows 40 minutes. Gumloop's comparison page says n8n agents need every tool wired in advance as sub-nodes. That is Gumloop's framing, not n8n's.

Apps, channels and where answers land

Each vendor counts its catalog in its own words. Gumloop offers 300+ built-in connectors plus any MCP server: a custom server from its URL on any plan, one hosted server on Pro, and proxied servers on Pro and Enterprise. A connector must be added on the agent itself, with a sign-in, before the agent can use it. n8n lists 500+ native integrations plus any API through its HTTP Request node, and n8n says it has 7,500+ community-built templates.

The bigger split is where people meet the work. A Gumloop agent can sit in a Slack channel, answer in Microsoft Teams public standard channels for Microsoft 365 work or school accounts, get its own email inbox on gumloop.ai, run on a hosted page where visitors sign in with a Gumloop account, or be called through the API and MCP. n8n offers hosted and embedded chat and webhook triggers on every plan, and control through its API and CLI on every paid plan and self-hosted edition (not during the Cloud trial). For Gumloop's REST endpoints, SDKs and webhook URLs, see our Gumloop API guide. One limit matters for Gumloop: inbox senders and hosted-page visitors need Gumloop accounts, so it suits internal teams more than public chat.

Admin controls and where each side gates them

Gumloop Pro covers unlimited seats and teams on one credit pool, built-in organization roles (Admin, Manager, Member, Developer), shared team credentials and approval rules scoped to each agent. Enterprise adds SSO with SCIM and SAML, custom roles, audit logs, organization-wide connector policies, data exports, incognito mode and per-member credit limits. One pool, one gate. Without Enterprise, nobody can cap a single colleague's spending. On Pro and Pro trials, Gumloop's Terms let it use your content to train its own models; Enterprise is excluded.

n8n puts many of the same controls at its top tier. Both gate SSO high. Users are unlimited on every plan, and Cloud Pro adds admin roles and five days of workflow history. On n8n Cloud, SSO and LDAP, custom roles, log streaming, an external secret store, Git version control and execution data redaction are Enterprise features, while self-hosted Business includes SAML and LDAP. Teams comparing compliance should read both top tiers side by side, since each vendor quotes Enterprise privately. For other options beyond these two, see our Gumloop alternatives.

Matching the meter to the team that will own it

The choice follows who will maintain the system. A business team that wants agents answering in Slack, Teams or an email inbox, with a model picked per agent and no server to patch, fits Gumloop's hosted credit pool, provided someone watches usage, because cost per run varies and Pro stops at its cap. A technical team that wants to self-host, write code steps, keep data on its own servers and count whole runs per execution, with model calls billed separately, fits n8n, accepting that its agents are still labeled Preview. Run your real tasks through both trials before committing, keeping in mind that Gumloop's needs a card and turns into paid Pro unless you cancel before it ends.