Business-process automation
Connecting inboxes, spreadsheets, CRMs, databases, forms, messaging, and documents into a governed recurring process.
Independent tool overview
Gumloop at a glance
Gumloop is a no-code platform for building AI agents and deterministic workflows that use business apps, company data, web tools, models, code, and APIs. It supports conversational agents, subagents, reusable skills, scheduled and event triggers, visual workflows, 300-plus connectors, MCP servers, REST APIs, and SDKs. The platform can move from prototype to real operations quickly, but its agents and custom MCP tools can also take consequential actions. Production use needs least-privilege credentials, narrow tool rules, approval steps, idempotent workflows, failure alerts, cost caps, and independent output checks.
Visit the official Gumloop site ↗
- Product type
- No-code AI agent and workflow automation platform
- Current status
- Active
- Core builders
- Conversational agents and visual workflows
- Integrations
- 300+ connectors advertised; custom MCP supported
- Automation
- Schedules, webhooks, app events, API, and SDKs
- Models
- 35+ on Pro; custom proxy support on Enterprise
- Pro price
- Starts at $37/month with 20,000 credits
- Pro orchestration fee
- 8%
- Pro concurrency
- 5 workflow runs and 25 agent chats
- Enterprise
- Custom pricing and governance
- BYOK
- Pro and Enterprise
- Reviewed
- August 31, 2026
Overview
What Gumloop is
Gumloop has evolved from a visual AI-workflow builder into a broader agent platform. Teams can create role-specific agents that reason across tools and data, or build structured node-based workflows when the sequence and cost need to be predictable.
Agents can use app integrations, web search and browsing, a code and file sandbox, image generation, workflows as tools, custom or hosted MCP servers, and approved subagents. They can run interactively, on schedules, after external events, or inside workflows through an Agent node.
The visual workflow layer remains useful for repeatable transformations, enrichment, scraping, routing, notifications, and API-style pipelines. Native logic and many app nodes cost no extra credits, while AI, enrichment, web, custom-code, MCP, and agent work consume credits.
The current Pro plan starts at $37 per month with 20,000 credits, unlimited seats and agents, an 8% orchestration fee, 25 concurrent agent chats, five concurrent workflow runs, BYOK, collaboration, and agent-scoped connector guardrails. Enterprise adds negotiated capacity and organization-wide governance.
Gumloop is best viewed as an automation runtime, not a magic employee. Models can hallucinate, websites and documents can inject instructions, third-party schemas can change, and a technically successful run can still create the wrong external outcome.
Use cases
Who Gumloop is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Role-specific AI agents
Creating narrowly scoped research, sales, marketing, support, recruiting, operations, or analysis agents with defined tools and boundaries.
Mixed deterministic and agentic work
Using fixed workflow steps for retrieval, validation, and writes while reserving model judgment for classification, drafting, or exception handling.
Non-developer builders
Letting subject-matter experts model workflows visually and test them without building an orchestration service from scratch.
API and webhook pipelines
Triggering workflows from external products and retrieving structured outputs through webhooks, REST APIs, or the JavaScript and Python SDKs.
Teams needing model choice
Selecting among many hosted models or supplying provider API keys on paid plans to balance capability, cost, and governance.
Capabilities
Core Gumloop features
Custom agents
Configure a role, instructions, model, knowledge, tools, credentials, skills, triggers, and sharing policy for a reusable AI teammate.
Subagents
A parent agent can clone itself for parallel work or invoke explicitly allowed specialist agents in the same project.
Agent tools
Agents can use business apps, web and X search, browsing, files, code, images, workflows, and MCP servers according to configuration.
Skills and company context
Teams can capture reusable playbooks and connect shared company knowledge for agents and humans.
Visual workflow builder
Connect input, logic, loops, AI, scraping, enrichment, code, integrations, and output nodes on a canvas.
Triggers
Start agents or workflows on schedules, webhooks, email, Slack, Teams, forms, database changes, calendar events, or other supported signals.
Agent node
Run a configured agent inside a structured workflow and continue its conversation when the workflow provides a conversation ID.
MCP support
Use prebuilt, hosted, proxied, or custom Model Context Protocol servers to expose additional tools to agents and AI nodes.
Model selection and BYOK
Choose among hosted models and, on Pro or Enterprise, connect supported provider keys to reduce Gumloop's AI credit charge.
Credential modes
Agents normally use the runner's personal credentials, with configuration options for personal, pinned, team, or organization accounts where allowed.
App Rules
Pro and Enterprise users can block or tag matching agent tool calls; Pro supports up to three rules per agent and Enterprise adds organization-wide controls.
Run logs
Workflow history exposes status, node timing, credit cost, and the inputs and outputs for each executed node.
Alerts and retries
Pro supports failure email alerts, while time triggers can be configured with a maximum failure count.
Sharing and interfaces
Share agents and workflows with editor, viewer, use-only, organization, or link-based access according to the item's options and plan.
Enterprise governance
Adds custom roles, app policies, AI model controls, audit logs, SSO and SCIM, data exports, custom retention, and optional private infrastructure.
Process
How the Gumloop workflow works
Step 1
Pick one measurable outcome
Define the owner, trigger, source systems, expected artifact or state change, service level, failure cost, and success metric before choosing an agent or workflow.
Step 2
Use deterministic steps first
Model retrieval, filtering, joins, validation, routing, and writes as explicit workflow nodes. Add an AI or agent step only where judgment or unstructured input is genuinely needed.
Step 3
Design least-privilege identities
Use dedicated accounts and the narrowest OAuth scopes. Separate read, draft, approve, and execute authority, and do not give a general agent personal admin or production-owner credentials.
Step 4
Constrain the agent
Start with two or three tools, precise instructions, structured inputs and outputs, forbidden actions, a cost and step budget, and an explicit list of operations that require human approval.
Step 5
Treat inputs as untrusted
Email, Slack, websites, files, CRM notes, tickets, and MCP output can contain prompt injection. Never let retrieved content add tools, reveal secrets, change policies, or authorize an external action.
Step 6
Build an approval gate
Keep sending, publishing, purchasing, deletion, permission changes, financial commitments, customer contact, employment decisions, and production changes behind a deterministic review step.
Step 7
Make writes idempotent
Use stable event IDs, deduplication, upserts, existing-state checks, locks where needed, and a record of the intended change so retries cannot create duplicate messages, records, charges, or tickets.
Step 8
Test each component
Verify credentials and every tool or workflow independently, then test the integrated flow with normal, missing, malformed, duplicate, stale, adversarial, high-volume, timeout, and provider-error inputs.
Step 9
Inspect the run log
Confirm which nodes and tools ran, their inputs and outputs, external side effects, latency, credit cost, failures, and final result rather than trusting the agent's summary.
Step 10
Calibrate model and credits
Run a representative sample, compare budget and advanced models, reduce unnecessary tools and context, batch deterministic work, and set usage notifications and overage limits.
Step 11
Release gradually
Begin with read-only recommendations or drafts, shadow the human process, approve each result, then expand autonomy only after measured accuracy and failure handling are acceptable.
Step 12
Operate it like production
Add failure alerts, freshness checks, retry limits, dead-letter review, audit evidence, dashboards, an emergency disable path, credential rotation, versioned prompts and workflows, and a documented human fallback.
Cost
Gumloop pricing and free plan
Gumloop uses subscription credits plus an orchestration fee. Pro starts at $37 per month with 20,000 credits and an 8% orchestration fee; additional plan sizes are available. Credits pay for model, tool, enrichment, web, custom-code, MCP, and workflow execution according to the node or agent path. Overage is documented at $0.007 per credit up to twice the monthly allocation. Enterprise pricing and discounts are negotiated.
Free
$0/month
A limited account is still referenced throughout current documentation; Gumloop's last public allocation announcement listed 5,000 credits per month.
- 2 concurrent workflow runs
- 5 concurrent agent interactions
- One free trigger was introduced in 2025
- Verify current signup allowance because the live pricing table focuses on Pro and Enterprise
Pro
Starts at $37/month
Individual and collaborative paid plan with 20,000 included credits at the entry level.
- Unlimited seats and agents
- 35+ models
- Bring your own API keys
- 8% orchestration fee
- 5 concurrent workflow runs
- 25 concurrent agent chats
- Agent-scoped connector rules
- 14-day trial advertised
Pro overage
$0.007/credit
Optional usage beyond the included monthly credit balance.
- Documented cap of 2x monthly allocation
- Credit notifications are configurable
- Unused non-Enterprise credits do not roll over
- Tool and workflow costs still apply
BYOK on Pro
Provider charges plus Gumloop credits
Connect a supported model-provider key to reduce Gumloop's AI-model credit component.
- Workflow AI nodes drop to 1 Gumloop credit per call
- Agent model credits are reduced by 50%
- Tool credits and workflow base costs are unchanged
- The model provider bills its own usage separately
Enterprise
Custom pricing
Negotiated capacity, support, security, governance, retention, and infrastructure.
- Custom credits and concurrency
- Organization-wide connector and model controls
- RBAC, SAML and SCIM
- Audit logs and data exports
- Custom retention
- Optional VPC and embedded expert
- Orchestration discounts may be available
Pricing checked . Check current pricing at the source ↗
Assessment
Gumloop strengths and limitations
Where it stands out
- Combines conversational agents and deterministic visual workflows in one platform.
- Broad connector coverage and MCP support reach many common business systems.
- Subject-matter experts can build useful automations without owning a custom orchestration stack.
- Agents can invoke focused workflows instead of improvising every business step.
- Subagents allow parallel research or delegation to explicitly selected specialists.
- Schedules, event triggers, webhooks, APIs, and SDKs cover interactive and unattended use.
- Run logs expose per-node inputs, outputs, timing, and credit consumption for debugging.
- Native logic and many integration nodes cost zero credits, making deterministic workflow steps economical.
- Multiple model choices and BYOK let teams tune cost, behavior, and vendor selection.
- Personal-credential execution can preserve each runner's source-system permissions.
- Pro includes unlimited seats, making collaboration less dependent on per-seat pricing.
- Agent-scoped rules on Pro provide a practical first layer of tool guardrails.
- Enterprise offers a substantial governance set including roles, app policies, model controls, audit logs, SSO, and exports.
- Official documentation is unusually detailed about costs, credentials, concurrency, troubleshooting, and production controls.
What to consider
- Agents can hallucinate facts, misunderstand intent, choose the wrong tool, generate invalid arguments, and claim completion when an external action failed.
- Any connected email, document, website, CRM record, ticket, message, or tool result can contain prompt injection or manipulated evidence.
- Custom MCP tools are immediately available once exposed, and Gumloop's custom-MCP guide says agent and Ask AI uses do not provide per-call approval prompts.
- Data retrieved from one MCP server can be passed to another, creating cross-system exfiltration risk unless tools and credentials are tightly scoped.
- Pro App Rules are limited to three per agent and apply only to that agent; organization-wide and custom-role rules require Enterprise.
- The custom-role documentation uses additive or least-restrictive composition in several places, so adding a role can unexpectedly broaden access or usage caps.
- Workflows and agents depend on third-party schemas, OAuth grants, website layouts, model behavior, and provider availability that can change independently.
- Agent costs vary with model, context, tools, reasoning steps, subagents, and invoked workflows, so the same apparent task can consume different credits.
- The 8% Pro orchestration fee, Gumloop credits, provider charges under BYOK, and paid external services all affect total cost.
- Failed workflows still charge for every node that executed before the failure.
- Loops, enrichment nodes, web agents, subagents, and long conversations can increase credit consumption rapidly.
- At the entry Pro limits, only five workflow runs and 25 agent chats can be active concurrently; Free is lower and rejects work at its limit.
- Alerts are a Pro feature, while many of the strongest audit, role, retention, and data-export controls require Enterprise.
- A shared or pinned credential can let a workflow act with more authority than the person who initiated it.
- Anyone-with-link and public sharing require careful review because configuration, data, or executable behavior may reach unintended users.
- Run logs can contain sensitive inputs and outputs, so access, retention, and export policy matter as much as workflow permissions.
- The live pricing page does not display the Free tier or its current credit allowance, so prospective users should verify the signup screen rather than rely on the last blog announcement.
- No-code does not remove the need for data contracts, retries, idempotency, test cases, monitoring, incident response, and change management.
- Gumloop should not autonomously make medical, legal, financial, employment, identity, security, or safety-critical decisions.
- A production automation still needs an accountable owner and a documented manual fallback.
Compare
Gumloop alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Agents
Zapier Agents
Choose Zapier Agents when Zapier's large automation ecosystem and agent experience better match existing business workflows.
Explore Zapier Agents →Business Operations
Zapier
Choose classic Zapier when deterministic trigger-action automation and broad app coverage matter more than multi-step autonomous agents.
Explore Zapier →Business Operations
n8n
Choose n8n when technical teams want more deployment control, code-level flexibility, and self-hosting options.
Explore n8n →Business Operations
Lindy
Choose Lindy when the priority is quickly configuring business-facing AI assistants and recurring operational tasks.
Explore Lindy →Business Operations
Bardeen
Choose Bardeen when browser-centered productivity and sales workflows are more important than a general agent and workflow runtime.
Explore Bardeen →Questions
Gumloop FAQs
What is Gumloop?
Gumloop is a no-code platform for building AI agents and visual workflows that use business apps, data, models, web tools, code, triggers, APIs, and MCP servers.
Is Gumloop still a workflow builder?
Yes. Its current positioning emphasizes agents, but the node-based workflow builder remains available for structured, repeatable automations and for tools an agent can call.
How much does Gumloop cost?
Pro currently starts at $37 per month with 20,000 credits and an 8% orchestration fee. Enterprise is custom priced. Current docs also reference a limited Free account, although the live pricing table does not display its allowance.
What is a Gumloop credit?
Credits are Gumloop's usage currency. Workflow base runs, AI models, tools, enrichment, web operations, custom code, MCP, agent reasoning, and invoked workflows can each contribute to the total.
Can I use my own model API key?
Yes, on Pro or Enterprise. Current docs say BYOK reduces workflow AI nodes to one Gumloop credit and cuts agent AI-model credits by 50%, while the provider bills its own usage.
Can Gumloop agents run automatically?
Yes. Agents can use schedules and supported external events, and an Agent node can run inside a workflow triggered by a schedule, webhook, app event, API, or SDK.
Does Gumloop support MCP?
Yes. It supports prebuilt and custom MCP servers, with hosted and proxied MCP capabilities for eligible organizations.
Are custom MCP tool calls approved individually?
The current custom-MCP guide says approval prompts are not available for agent or Ask AI use. Expose only narrow tools and credentials, and place consequential actions behind a separate deterministic approval workflow.
How do credentials work for shared agents?
By default, agents use the credentials of the person running them. Builders can also configure pinned, team, or organization credentials where permitted, which requires careful authority review.
Can agents invoke other agents?
Yes. An agent can clone itself for parallel work and can call specifically added custom agents in the same project as subagents.
How can I control agent actions?
Limit tools and credentials, write explicit approval rules, use agent-level App Rules on Pro, use organization governance on Enterprise, and route consequential writes through deterministic reviewed workflows.
What happens when a workflow fails?
The run log shows the failing node and its inputs and outputs. Gumloop charges for nodes that ran before the failure, so production flows also need retry limits, idempotency, alerts, and a recovery queue.
Is Gumloop appropriate for sensitive company data?
It can be evaluated for sensitive work, but the answer depends on plan, contracts, retention, connected model providers, credentials, logs, regions, roles, and required controls. Enterprise offers the broadest governance set.
Can Gumloop replace an operations team?
It can automate bounded repeatable work, but human owners are still required for policy, exceptions, high-impact decisions, quality review, incidents, and vendor or source-system changes.
Bottom line
Our Gumloop verdict
Gumloop is one of the more complete no-code options for teams that want both AI agents and conventional workflows. Its connector reach, subagents, triggers, MCP support, model choice, detailed run logs, and low entry Pro price are compelling. The deciding question is governance, not whether a demo works. Use deterministic nodes for business rules, give agents only narrow tools and identities, build explicit approval gates around writes, test retries and hostile inputs, and model all three cost layers—subscription, credits or orchestration, and external providers—before scaling.
Visit Gumloop website ↗Similar Business Operations Tools
View all Business Operations AI tools →
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.