Quick answer: what Gumloop costs in 2026
Gumloop Pro costs $37 a month and includes 20,000 credits, after a 14-day free trial of Pro that requires a card. Enterprise is custom-priced through sales, with no published figure. One credit is $0.005. Usage past the monthly allotment bills at $0.005 per credit once pay as you go is enabled.
Every number here comes from Gumloop's pricing page and its credits documentation on September 10, 2026, with no third-party estimates. One caveat: the page carries structured data listing a $0 "Free plan." But its description is only the free 14-day trial, as shown on the vendor page in September 2026. There is no permanent free tier.
Gumloop pricing plans at a glance
Two paid tiers, plus a metered extension you switch on yourself:
| Plan | Price | Credits | Notes |
|---|---|---|---|
| Pro | $37 a month | 20,000 a month | 8% orchestration fee |
| Enterprise | Custom, via sales | Custom, with rollover | Fee discounts, VPC |
| Overage on Pro | $0.005 per credit | Capped, 1,000,000 max | Off by default |
Overage is not a third plan. It is a switch on Pro that keeps agents running past the monthly grant, and it stays off until you turn it on.
What a Gumloop credit is worth
Gumloop meters everything in credits, and the conversion is fixed at $0.005 per credit. That makes the arithmetic work in both directions. A dollar buys 200 credits. A model call that lists at $0.03 shows up as 6 credits. Charges round up to whole credits, so nothing lands as a fraction on the log.
"Agent credit costs are variable. The same agent might cost 2 credits for a quick question and 200 for a deep research task. You pay for what the agent actually does, not a fixed per-message price. ... 1 credit = $0.005 ... $1 buys 200 credits. ... Divide the model cost by $0.005: a $0.03 model call is 6 credits. ... Charges are rounded up to whole credits."
Source: docs.gumloop.com
Not all of the 20,000 credits on Pro are bought at that indexed rate. Pro includes 7,400 credits at list price plus 12,600 bonus credits, so the allotment is worth more than $37 of list-price usage. The bonus is attached to the plan, not to headcount.
What none of this tells you is what one agent run costs, because that figure is genuinely variable. Gumloop's own examples span two orders of magnitude: a quick question can cost 2 credits, a deep research task 200. Three things drive the Chat and Reasoning component. The model you pick matters, and so does the length of the conversation, since every message carries the earlier exchange as context. So does the number of tools connected, since each tool adds its definition to the prompt. One message can also trigger several billed AI steps when the agent loops through tools. Budget from your own logs, not from one sample run.
How a run is billed: four components
An agent chat is billed as the sum of four line items. Read them separately before you look at the total.
Chat and Reasoning is the model's own cost, converted into credits at $0.005 each. Tool Calls are 1 credit for each successful call, plus whatever the tool itself charges. Compute runs at 5 credits per session-minute of active processing, with a minimum of 1 credit per response. The Orchestration Fee is 8% of the other three on Pro.
Several rules cut the other way. Failed tool calls are not charged. Reading and writing ordinary apps such as Slack, Google Sheets and Gmail costs the 1 base credit and nothing more. Data tools such as Apollo enrichment or Firecrawl scraping add their own charge on top. Time the agent spends waiting for you costs nothing, so a conversation left open overnight does not drain the allotment. Enterprise customers running Gumloop inside their own virtual private cloud are not charged for compute.
Gumloop publishes a worked example worth copying into your model. A chat spends 400 credits on Chat and Reasoning, 10 on Compute and 15 on Tool Calls. That is 425 before the fee. The 8% orchestration fee on that comes to 34 credits after rounding up, and the chat closes at 459 credits. Note where the fee sits. It applies to the whole task, not to the model bill alone, so tool-heavy chats carry it too. Enterprise plans can negotiate a different orchestration fee, and the pricing page lists discounts on that tier rather than a rate.
Bring your own keys: what changes
Gumloop supports bringing your own API keys, and the trade is explicit. Point a model provider key at your account on Pro or higher, and Chat and Reasoning drops to 0 credits. You are paying that provider directly. In exchange, the orchestration fee rises from 8% to 16%, computed on the full list cost the task would have had without BYOK. You do not leave the meter. You change what it charges for.
Bringing your own key for MCP tool calls works differently. It reduces the credit cost to a minimum of 1 credit per tool call.
Whether the swap pays off depends on your own provider rates. Run it on real numbers instead of a rule of thumb. The question is narrow. Does the extra 8 points of fee, charged on the full list cost of every task, cost you less than the token credits you stop paying? Teams with committed model spend have the most to gain from running that comparison.
Free trial, rollover and overage rules
Every new account starts on a 14-day free trial of Pro. It requires a card, it is a one-time offer per customer, and it rolls into a paid Pro subscription unless it is canceled first. Treat it as a paid plan with a delayed first invoice.
"No. The 20k credit allotment is a monthly grant that resets each billing period. You get the full 20k again next month; you don't bank what you didn't use."
Source: gumloop.com
Credits do not roll over. The 20,000 on Pro are a monthly grant that resets each billing period, and whatever you did not spend is gone. Enterprise is the exception, with rollover listed on that tier.
Spend past 20,000 credits mid-month and agents stop working until pay as you go is enabled. That switch is off by default, which is reassuring for finance and inconvenient mid-task. Once it is on, from the Usage and Limits page, usage past the monthly credits bills at $0.005 per credit.
Overage on Pro is always capped. The default ceiling is 1,000,000 overage credits per billing period, which is $5,000, and the cap can be set lower. When the cap is reached, agents stop. Enterprise plans set their own cap or run uncapped. For most Pro teams the sensible move is to lower that ceiling to a number the budget can absorb before enabling overage.
What Pro and Enterprise include
The feature list below is as shown on the vendor page in September 2026. That pricing table is flattened text, so column attribution follows the page layout, Pro first and Enterprise second.
Both tiers list unlimited agents, 35+ models, bring-your-own API keys, unlimited seats and teams, shared credentials, Gumloop MCP and Gumloop CLI. Enterprise adds custom proxy support to the model list. Pro lists 20k credits a month, the 8% orchestration fee, 1 hosted MCP server, 5 concurrent workflow runs and 25 concurrent agent chats. Enterprise lists custom credits with rollover, discounts on the fee, custom MCP hosting and custom concurrency. Workflows appear under a Legacy label, which matters if you started your evaluation there.
Everything else in the Enterprise column is governance, not capacity. It covers dedicated Slack support, an optional embedded Gumloop expert, an optional virtual private cloud, org-wide connector policies, AI spend insights and a usage analytics agent. The rest is role-based access control, SCIM/SAML, an admin dashboard, audit logs, custom data retention rules, regular security reports, data exports, incognito mode and AI model access control. If procurement needs any of those, the route is a sales conversation.
The site banner also says the company raised a $50M Series B led by Benchmark. That is the vendor's own statement.
Where to see what a chat cost
The running cost of a conversation appears live in the chat header. A details drawer lists one row per credit type charged. The rows are Chat and Reasoning, Tool Calls, Compute, Orchestration Fee, Evaluation and Self-Improvement, Subagents, Model Routing when the agent is set to Auto, and Other. Waived credits still appear, struck through and labeled Free, so BYOK and VPC savings stay visible.
Credit logs under Usage and Limits offer a grouped view, one row per conversation, and a detailed view, one row per charge. Organization admins see usage across all users. Logs export as CSV from Data Export, and the Insights dashboard is Enterprise only.
Gumloop's own advice for spending less: start with the Recommended model preset and reach for Auto or Smartest only when the task calls for it. Bring your own key so Chat and Reasoning drops to 0 credits on Pro or higher. Start fresh conversations instead of carrying old context. And stay inside the short context window, since some models charge higher long-context rates for the whole request.
How Gumloop's pricing compares with automation platforms
Headline prices are the wrong comparison. What matters is what each meter counts.
Gumloop counts the model bill: tokens at cost, tool calls, compute, plus 8%. Zapier counts tasks, meaning anything the platform successfully completes on your behalf, such as sending an email or creating a spreadsheet row. Its Professional plan is $19.99 per month billed annually for 750 tasks a month. The free plan is $0 a month for 100 tasks across Zap workflows, Tables and Forms.
Make counts credits consumed by each module action in a scenario. Core starts at $9 per month billed annually for 10,000 credits a month, and a free plan covers up to 1,000 credits. Lindy counts per-user credits. The Plus plan is $29.99 a month per user with 3,000 credits per user per month, pooled across the workspace so every seat adds its allocation to one shared pool. Admins set allocations per seat, and Lindy pauses credit-using actions when the pool runs low rather than issuing a surprise bill.
Latenode counts runtime in CPU seconds rather than operations or steps, so workflow steps do not multiply the bill. The first 10,000 CPU seconds are free every month on both plans, and the rate is $0.00012 per CPU second in the first paid bracket.
That difference decides who pays what. An agent-heavy team pays mostly for tokens on Gumloop, and mostly for volume or seats elsewhere. A team running many small app-to-app jobs pays per completed action on a task or credit meter, and per second on a runtime meter. None of these units convert into each other, so test them against your own month.
Who Gumloop's pricing suits and who should look elsewhere
Gumloop's pricing suits teams that want agents with a predictable entry cost and a hard stop on the downside. Pro is $37 a month for 20,000 credits. Overage stays off until you enable it, and the cap can be set below the 1,000,000-credit default. It also suits teams already paying for model keys, since BYOK takes Chat and Reasoning to 0 credits and leaves the 16% orchestration fee as the platform's cut.
It fits less well in two situations. First, when the work is thousands of simple app-to-app runs a month. A 1-credit tool call, plus model tokens, plus the fee, adds up on volume that a task or runtime meter prices differently. Second, when a permanent free tier is a requirement, because the pricing page offers a 14-day trial of Pro and paid usage after it. Model your own month first. The gap between a 2-credit question and a 200-credit research task is too wide for a rule of thumb.
