Gumloop Pricing in 2026: Plans, Costs, and Limits Explained

Pricing checked September 26, 2026. USD; calculations exclude taxes and negotiated discounts.

Gumloop publishes two plans: Pro starting at $37 per month with 20,000 included credits, and Enterprise on a quote. There is no free tier — only a 14-day trial. Both plans include unlimited agents, unlimited seats and unlimited teams. Gumloop pricing

The subscription is the small part of the bill. Gumloop charges for what your agents actually consume, one credit is worth $0.005, and an 8% orchestration fee sits on top of every run. Whether that adds up to $37 a month or several hundred depends entirely on how model-heavy your agents are.

This guide breaks the credit formula into its four components, works through a lead-enrichment agent at two volumes, and explains the one setting you should change on day one.

Gumloop pricing at a glance

Pro allows 5 concurrent workflow runs and 25 concurrent agent chats; Enterprise is custom on both. Note that Gumloop now lists workflows under "Workflows (Legacy)" — the product's centre of gravity has moved to agents, which is worth knowing before building a long-term automation estate on the older model. Official plan comparison

How Gumloop calculates your bill

One credit equals $0.005, so $1 buys 200 credits. An agent run's cost has four components: Gumloop credits documentation

Gumloop's own worked example: a chat spending 400 credits on Chat & Reasoning, 10 on Compute and 15 on Tool Calls pays a 34-credit fee — 8% of 425, rounded up — for 459 credits total. At $0.005 that single chat costs about $2.30.

Compute is worth translating: 5 credits per session-minute is $0.025 a minute, or $1.50 per hour of active processing. Waiting for a user to reply is not charged as compute.

The BYOK trade-off, which runs counter to intuition

Bringing your own API key raises the orchestration fee from 8% to 16%. That sounds like a penalty, but Chat & Reasoning then costs zero credits — you pay your model provider directly instead. The 16% fee is calculated on the run's value before the model-credit waiver, not just its remaining compute and tool charges.

Using the figures above: on Gumloop's models, 8% of 425 credits is 34. On BYOK, model credits are waived, but the fee remains based on 425 credits: 16% is 68. Add 25 compute/tool credits for a total of 93, plus the model provider's separate bill. BYOK is therefore cheaper on Gumloop credits whenever model reasoning dominates the run, which for most agents it does. The question becomes whether your negotiated model rates beat Gumloop's $0.005-per-credit conversion.

Credits do not roll over, and overage is expensive

Credits reset monthly on Pro with no rollover; Enterprise can negotiate rollover. To run past 20,000 credits you must enable overage on the Usage & Limits page, billed at $0.005 per credit.

Compare the two rates. Included credits work out at $37 ÷ 20,000 = about $0.00185 each. Overage costs $0.005 each — roughly 2.7× more. Overage is not a smooth extension of your plan; it is a materially worse rate, so size your usage to stay inside the allowance where you can.

What does Gumloop cost for a real agent?

Illustrative calculations using the documented credit formula, with assumed per-run consumption. Not invoices. Substitute your own measured figures.

Example 1: lead enrichment at 100 leads per month

An agent reads a new lead, enriches it from an external source, scores it and writes back to the CRM.

Example 2: the same agent at 300 leads per month

Tripling the volume multiplied the bill by 5.7×, not 3×, because everything above 20,000 credits is charged at the higher overage rate. This non-linearity is the single most important thing to model before rolling an agent out more widely.

The same 300 leads on BYOK

With your own key, Chat & Reasoning leaves the credit bill: 18 credits of Tool Calls and Compute, plus a 16% fee on the original 168-credit value (26.88, rounded up to 27), is 45 credits per lead. At 300 leads that is 13,500 credits — comfortably inside the allowance, so Gumloop costs $37 and your model provider bills you separately for the reasoning.

Whether that is cheaper overall depends on your model rates versus Gumloop's conversion. It is clearly worth testing at this volume, since it is the difference between $37 and $210 on the Gumloop side.

Change the overage cap before you ship

Pro's overage cap defaults to 1,000,000 credits per billing period — $5,000. You can lower it, and you should, before an agent loop or a retry storm discovers that ceiling for you. Set it to a number you would be comfortable seeing on an invoice.

Where Jet Admin fits alongside Gumloop

Gumloop runs AI agents; Jet Admin gives people an interface for the data those agents touch. These are complementary, and their credits are not the same unit — Gumloop credits meter agent execution, Jet Admin credits meter AI consumption during app building. Do not convert between them.

The pairing matters most where an agent should not act unsupervised. An enrichment agent can score leads, but somebody usually reviews the borderline ones, and a reviewed queue with clear permissions is a different product from an agent runtime.

Sources: Gumloop pricing, Gumloop credits, Jet Admin pricing.

There is one real cost interaction, and it is testable rather than assumed. Because Gumloop bills compute per session-minute and reasoning per token, work that does not need an agent is expensive to run through one. Routing, lookups and record display are cheaper as ordinary application logic. If your agent is spending reasoning credits deciding which record to show a human, that part belongs in an interface — measure the credit reduction rather than assuming it.

When to keep Gumloop as-is

For genuinely agentic work — research, unstructured triage, tasks where the sequence is not knowable in advance — Gumloop's model access, MCP support and BYOK option are the point, and no application layer substitutes for that.

Which plan should you choose?

Pro is the only self-service option, so the real decision is how you configure it: measure a representative agent run first, enable BYOK if reasoning dominates your credits, and lower the overage cap immediately.

Move to Enterprise when you need RBAC, SCIM/SAML, audit logs, VPC deployment or credit rollover — or when your steady-state overage is large enough that a negotiated orchestration fee and custom credit bundle beat paying $0.005 per credit.

If your monthly overage is consistently exceeding the subscription several times over, that is the signal to negotiate rather than keep paying list rate.

Frequently asked questions

Does Gumloop have a free plan?

No permanent free tier is listed. The 14-day Pro trial requires a card and becomes a paid subscription unless cancelled; Pro starts at $37/month.

What is a Gumloop credit worth?

One credit is $0.005, so $1 buys 200 credits. Overage is billed at that same $0.005 per credit.

What is the 8% orchestration fee charged on?

The combined total of Chat & Reasoning, Tool Calls and Compute for a run. It rises to 16% when you bring your own API key — model reasoning is waived in Gumloop credits, but the fee uses the run's pre-waiver value and the provider bills tokens separately. Compare the combined total.

Do Gumloop credits roll over?

Not on Pro. Enterprise plans can have unused credits roll over.

What happens if I exceed 20,000 credits?

Agents cannot continue past the included allowance without available credits or enabled overage on the Usage & Limits page. Once enabled, extra credits cost $0.005 each, with a default Pro cap of 1,000,000 credits ($5,000) that you can and should lower.

How much does Gumloop Enterprise cost?

Not published. Request a quote covering credit volume, the orchestration fee, rollover terms, deployment and security requirements.

Measure one agent run before you roll out

Run your agent ten times, read the credit breakdown, and multiply by realistic monthly volume. Then check the result against 20,000 — because the step from inside the allowance to overage is a 2.7× rate change, not a gentle slope.

If that breakdown shows reasoning credits going into decisions a person should be making, put that step behind a proper review interface and re-measure what the agent still needs to do.