Gumloop Pricing: Pro at $37 a Month and How Credits Are Spent

Gumloop Pro costs $37 a month, and that single subscription covers the whole organization: unlimited agents, seats and teams drawing on one shared pool of 20,000 credits a month ^[1]. Enterprise is priced by quote. There is no free plan and no annual price. Each credit is worth $0.005, so the real question is how fast your agents spend the pool. That varies a lot. Gumloop's docs say the same agent can spend 2 credits on a quick question and 200 on a deep research task ^[2]. Pro suits a small team running agents on demand. Unwatched polling triggers are the risk.

What one Pro subscription includes

The pricing page prints starts at beside the Pro price, but no larger Pro option sits above it. There is one Pro price, $37 a month, and the only way to spend more on Pro is overage. Everything else on the page is a gate between Pro and Enterprise. The grid below holds the figures that decide the bill.

By our arithmetic, $37 a month at the list rate buys 7,400 credits, so the 20,000 credits included each month are worth $100 and the 12,600 bonus credits $63. That works out to $1.85 per 1,000 included credits a month. The bonus is the whole value case. It pays off only if your team spends the pool before it resets, because unused credits vanish each billing period. Seats cost nothing extra. A tenth teammate changes nobody's invoice; it only adds one more person drawing on the same 20,000 credits.

Where each credit goes

Every agent chat shows its cost lines in a details drawer. Chat and Reasoning bills model tokens at the provider's list price, thinking tokens included. Tool Calls charge at least 1 credit per successful connector call; tools that fetch outside data, such as Apollo or Firecrawl, add their own charge on top. Compute costs 5 credits per session-minute of active processing, and waiting for you is free. The orchestration fee then adds 8% of those three lines, rounded up. Brain searches appear as a fifth line when used.

Gumloop's docs walk through one run: 400 credits of model tokens, 10 of compute and 15 of tool calls add up to 425, the fee adds 34, and the total is 459 credits. By our arithmetic, that run costs about $2.30. The agent browser and code sandbox carry no per-call charge, only compute. At 5 credits a minute, 1,000 credits buy 200 session-minutes before the 8% fee, or 300 credits ($1.50) for an hour of active processing. Long browser sessions add up quietly.

A few meters sit outside that core. Some models offer faster speed modes at a higher token rate. Talk mode, the voice call feature, uses about 10 credits a minute, roughly $0.05 a minute before model, tool and compute charges. Live artifacts bill about 1 credit per 55 seconds, paid by whoever views them. Brain indexing charges on the first sync after you approve a free estimate, then only for new or changed documents.

The meter most likely to surprise you is the AI-created trigger. These custom triggers bill every poll, whether or not anything fires, at a minimum of 1 credit per poll. They poll every five minutes by default, the shortest interval, and can be slowed to as rarely as once a week. Left at five minutes, that means 288 polls a day. By our arithmetic, one such trigger spends at least 288 credits a day, about 8,640 credits in 30 days. That is roughly 43% of the Pro pool, before any decision-model calls. Poll hourly when you can.

Bringing your own model key: zero tokens, double fee

On Pro or higher you can connect your own key from OpenAI, Anthropic, Google AI, Perplexity, xAI or Fireworks AI. Model calls the key serves then cost 0 credits; tool calls and compute stay billed. The orchestration fee rises to 16%, computed on the full list cost the run would have had without the key. The 459-credit example drops to 93 credits, about $0.47 by our arithmetic. The token cost moves to your provider account. A key covers only models its provider serves.

That trade saves Gumloop credits on chat-heavy agents, where tokens dominate the bill. On runs that are mostly browser time or tool calls it can even add credits, because the 16% fee also applies to the compute and tool lines, while the tokens are still paid to your provider. A second kind of key matters for paid data tools: connecting your own Apollo account drops each call to the 1-credit minimum, and Gumloop's docs say DataForSEO through your own account costs zero Gumloop credits. On Pro, keys are personal or shared per team; Enterprise adds organization keys that override both.

When the 20,000 credits run out

The pool resets every billing period, and nothing rolls over except on Enterprise. Pay as you go is off by default; you turn it on from the Usage and Limits page. Overage bills at $0.005 per credit, which by our arithmetic is $5 per 1,000 credits. On Pro, overage is always capped: the default ceiling is 1,000,000 credits, or $5,000 a billing period on top of the $37 a month, and you can lower it ^[2]. Enterprise sets its own cap or runs uncapped.

When the pool or the cap runs dry, work stops. Agents stop answering until pay as you go is on, or, once the overage cap is reached, until you raise the cap or the billing period resets. Trigger runs fail and are re-enabled once credits return. Brain searches stop, while a running agent continues without them. Talk mode calls end, and evaluations are skipped silently. Gumloop sends an out-of-credits email by default, plus usage emails at 75% and 90% of the pool, and both thresholds can be changed.

Web chat, Slack and API usage all draw from the same organization pool, and an agent's owner gets no separate bill. Someone outside the organization using a shared agent pays from their own account. A hosted chat page charges the person chatting. What Pro lacks is a per-person ceiling. Monthly credit limits per member come through custom roles, an Enterprise feature. On Pro, one busy teammate can drain the pool for everyone.

Trial, billing and cancellation rules

There is no free plan. The page's structured data does list an offer called a free plan, but its description is the 14-day trial. That trial is one-time per customer, needs a card, and turns into paid Pro unless you cancel before it ends. Cancel from Manage Subscription on the Usage and Limits page, which opens the Stripe portal; in an organization only the Admin role can open it. Access runs to the trial's end without a charge. Our guide to whether Gumloop is free covers what costs nothing.

Gumloop's pricing page prints only a monthly Pro price; it shows no annual plan. Enterprise is invoiced by contract. Subscriptions renew automatically until canceled. Under Gumloop's Terms, paid subscription fees are non-refundable except where the law requires otherwise ^[3]. Canceling a paid plan keeps it and its credits usable to the end of the billing period. Agents, skills, Brain sources and chats stay in the account afterward, so you can resubscribe later.

How the credit model compares with n8n, Zapier and Lindy

These rivals bill on different units, so sticker prices do not line up. n8n charges per workflow execution, however many steps it takes: its Cloud Starter plan costs $20 a month billed annually or $24 billed monthly for 2,500 executions a month. Zapier counts tasks: its Professional plan costs $19.99 a month billed annually or $29.99 billed monthly for 750 tasks a month. Lindy bills every active user a monthly seat with credits; its pricing page opens at $29.99 per user a month for 3,000 credits, or on some visits $49.99 per user a month for 5,000 credits.

A per-execution or per-task count is easier to forecast, but on n8n the model's tokens are billed separately, through your own provider key or n8n's Gateway credits, while Gumloop's credits already include them. Gumloop's shared pool suits a team adding people, but each run's cost moves with the model, the conversation length and the tools it calls. Trials differ too. n8n Cloud and Zapier Professional both offer 14 days without a card, and Lindy gives credits for 7 days without one. Gumloop asks for a card, and its trial turns into paid Pro unless you cancel before it ends. Our buyer's checklist lists the controls to check before paying.

Will 20,000 shared credits cover your team?

For a small team chatting with agents on demand, the pool goes a long way: a quick question can cost 2 credits and a deep research task about 200, and even the docs' 459-credit worked example stays under 500. It shrinks fastest under AI triggers left at five minutes, long browser sessions and paid data tools. Lower the overage cap in the first month, watch credits per task on the Performance page, and slow down quiet triggers. Per-member limits or rollover mean an Enterprise conversation. The full Gumloop review weighs everything beyond price.