Gumloop Review 2026: Pricing, Features and Verdict

What it is

Gumloop started as a no code workflow builder and now describes itself as agent infrastructure for AI native companies. You give an agent a goal, connect the services it needs, add instructions and skills, and it decides which tools to call. Its homepage lists customers including Shopify, Instacart, Gusto, Ramp, Samsara and Verkada.

There are three main pieces:

• Agents. Goal driven assistants that call connectors, run code, browse the web and hand work to subagents.
• Workflows. The original node based canvas for fixed, repeatable pipelines, still available alongside agents.
• Gumball. A personal agent with chat, an inbox and reports that uses your connected accounts and can accept voice messages.

Agents live where your team already works. According to the docs, you can reach one in a Slack channel or DM, in Microsoft Teams, through a dedicated email inbox, on a hosted web page, through the API and SDK, or as an MCP server that other AI tools call.

Pricing explained

Gumloop has one self serve plan and one sales led plan.

Plan	Price	Credits	Key limits
Pro	$37 a month	20,000 a month	Unlimited agents, seats and teams; 35+ models; 5 concurrent workflow runs; 25 concurrent agent chats; 1 hosted MCP server; email support
Enterprise	Quote only	Custom, with rollover	Custom concurrency, orchestration fee discounts, VPC option, SAML and SCIM, RBAC, audit logs, admin dashboard, custom data retention, dedicated Slack support

The pricing page lists a monthly Pro price only; we did not find an annual price.

How a credit is calculated

According to Gumloop's credits documentation, one credit is $0.005, so $1 buys 200 credits. Every agent chat is billed on four parts:

1. Chat and reasoning. Model tokens at the provider's list price, converted to credits. A $0.03 model call is 6 credits.
2. Tool calls. 1 credit per successful connector call, plus that tool's own charge if it has one.
3. Compute. 5 credits per minute of active session time.
4. Orchestration fee. 8% of the three items above.

Gumloop's own example: 400 credits of reasoning, 10 of compute and 15 of tool calls is 425 credits, plus a 34 credit fee, for 459 credits. That is about $2.30 for one heavy research chat.

Worked examples

• Light chats. A quick task with a $0.03 model call (6 credits), two connector calls (2 credits) and one minute of compute (5 credits) is 13 credits, plus a 1 credit fee, so about 14 credits. 20,000 credits covers roughly 1,400 of those chats a month for $37.
• Heavy research chats. At 459 credits each, the same 20,000 credits covers about 43 chats.
• A team that needs 60,000 credits. $37 for the first 20,000 plus 40,000 overage credits at $0.005, which is $200, for $237 a month.

Overage is opt in. Without it, agents stop at the credit limit until the next period. With it, Pro overage is capped by default at 1,000,000 credits ($5,000) per billing period.

Bring your own key. If you connect your own model API key, eligible model calls cost 0 credits and you pay the provider directly, but the orchestration fee doubles to 16%. That suits teams that already have committed spend with OpenAI or Anthropic.

Hidden costs to watch. Tools with their own charge (enrichment and search providers) add credits on top of the 1 credit call fee. Long running browser sessions burn compute at 5 credits a minute. Chat costs show in the chat header and under Settings, then Organization, then Usage and Limits, so check them after your first week.

For how $0.005 a credit stacks up against other builders' units, see the AI Price Index.

Key features

Agents with approval controls

Each agent has a model, instructions, connectors and skills. You can set max steps from 1 to 200 (default 100), verbosity, parallel tool calls and output length. Every tool gets an approval rule: always allow, ask each time, ask only for writes and deletes, or a custom setting. That last option matters when an agent can send email or edit a CRM record.

35+ models with fallback

Pro includes 35+ models. Model fallback switches to another provider if the primary one is down, and the homepage describes a model router that matches tasks to a suitable model for cost. Enterprise adds custom proxy support and controls over which models staff can use.

400+ connectors and MCP

The homepage lists 400+ connectors, including Slack, HubSpot, Salesforce, Gmail and Notion. Pro can host one MCP server; Enterprise adds hosted and proxied MCP servers. That is fewer apps than Zapier's 9,000+ or Make's 3,500+, so check your long tail tools before you commit.

Subagents, skills and a code sandbox

Agents can delegate to specialist subagents or clone themselves, with up to 10 subagents running at once. Skills are reusable instructions, templates and scripts that agents can create and update themselves. A built in sandbox runs Python and shell with 80+ packages preinstalled, so an agent can clean a CSV or chart a result without a separate tool.

Real browser automation

According to the homepage, agents can drive a real browser with session recording and CAPTCHA handling, and the docs cover saved browser logins and profiles. That opens up sites with no API, at the cost of compute minutes.

Team features

Unlimited seats on Pro is the headline. Agents can be shared across teams, and the docs explain who is charged when a shared agent runs. Shared credentials and the Company Brain knowledge layer are listed under Enterprise, so on Pro each person typically connects their own accounts.

What users say

Review volume is thin for a platform with Gumloop's customer list. We found no G2, Capterra or Trustpilot listing with 10 or more reviews that we could open. Product Hunt shows 5.0 from 9 reviews, too few to weigh heavily.

Recurring praise on Product Hunt:

• Ease of use. Reviewers describe it as a more capable Zapier that non technical staff can learn.
• Range of problems. Users like that agents handle multi step research and ops tasks that fixed workflows cannot.
• Founder support. Several mention fast responses from the team.

The only improvement request there was about ease of use compared with Dify. Nine reviews cannot tell you how billing holds up at volume, so treat the 14 day trial as your real test and watch the per chat cost readout.

Privacy and security

• Certifications. SOC 2 Type II is listed on the homepage.
• Access. SAML SSO, SCIM and role based access control on Enterprise. Pro has agent scoped connector policies; org wide policies are Enterprise only.
• Data handling. Custom data retention, security reports and audit logs are Enterprise features. A VPC deployment option is available on Enterprise; there is no self hosted edition.
• Model training. The pages we opened do not state Gumloop's model training policy. Bring your own key routes model calls through your own provider contract, which is the cleanest way to apply your existing data terms.
• Secrets. The docs describe a code sandbox with managed secrets, so API keys are not pasted into prompts.

Who it is for

• Marketing and growth teams running research, enrichment and content agents where several people need access.
• Ops teams that want agents answering in Slack or Teams rather than a separate app.
• Startups that want 35+ models and a code sandbox for $37 a month without paying per seat.
• Companies with model contracts that prefer bring your own key billing.

Who should skip it

• Anyone who needs a free plan. Zapier, Make and Botpress all have permanent free tiers.
• High volume, fixed automations. A simple trigger and action pipeline is cheaper per run on Make or n8n.
• Teams that must self host. n8n's Community Edition runs on your servers; Gumloop offers VPC only on Enterprise.
• Buyers who need SSO on a small budget. SAML and SCIM require an Enterprise quote.
• Customer facing chat or voice agents. Gumloop is built for internal work; Botpress and Voiceflow are built for conversations with customers.

Alternatives

	Gumloop	Zapier	Make	Lindy	Relevance AI
Entry paid plan	$37 a month, 20,000 credits	$29.99 a month, 750 tasks	$16 a month, 10,000 credits	$49.99 a month, 5,000 credits	$29 a month, 2,500 Actions
Free plan	14 day trial	Yes	Yes	7 day trial	No, free plan retired
Self host	No, VPC on Enterprise	No	No	No	No

• Zapier review: the widest app catalog and Agents on the free plan.
• Make review: the cheapest per action for fixed automations.
• n8n review: self hosting and per execution billing for technical teams.
• Lindy review: an AI assistant for email, meetings and scheduling.
• Relevance AI review: specialist agents for sales and go to market teams.

Our best AI agent builders ranking places all of them side by side.

Verdict

Gumloop earns its slot as the best platform for teams building agents together. One $37 plan with unlimited seats, 20,000 credits and 35+ models undercuts per seat rivals as soon as three or more people need access, and the credit math is published down to the compute minute.

Go in with a usage plan. Run your real agents during the 14 day trial, read the per chat cost, multiply by monthly volume, and decide whether to turn on overage with a cap. If you need SSO, shared credentials or a free tier, price Enterprise or look at Zapier and Make first.