What Gumloop is
Gumloop combines two automation models. Visual workflows follow a defined graph of triggers, transformations, integrations, AI nodes, web tools, code, and outputs. Agents provide a conversational layer that chooses tools and can call workflows. This lets a team use deterministic processing where possible and reserve model discretion for variable work.
The distinction is important for both safety and cost. A workflow path can usually be inspected before it runs. An Agent can change its sequence, call different tools, and include growing conversation history. Do not use an Agent merely because the interface makes it convenient. A fixed data pipeline is easier to test and budget.
Workflows, Agents, scraping, and tools
Gumloop targets AI-heavy use cases such as web research, lead or company enrichment, content processing, CRM updates, and analysis. Web agents and scrapers can turn unstructured sites into inputs. Code and MCP components extend the platform beyond native connectors.
Each extension increases trust. A scraper must respect site terms, authentication boundaries, personal-data rules, robots guidance where applicable, rate limits, and intellectual-property constraints. Code should run with no unnecessary secrets or network access. MCP tools should be treated like installed software: review the server, actions, authentication, and update path.
Agents should receive a small tool set with precise descriptions. Separate read and write workflows. Require review before outreach, CRM overwrites, list exports, purchases, deletions, or publishing. A shared integration key can affect an entire organization, so it should not be available to every experimental Agent by default.
Credit pricing
Gumloop's credit documentation states that each workflow run starts with one base credit plus node costs. Many logic, formatting, and native integration nodes are free beyond that base. AI nodes were documented at roughly two credits for standard models, 20 for advanced models, and 30 for expert models. Web, enrichment, custom, and MCP nodes have their own rates.
Agent cost is variable. It depends on model, message length, accumulated conversation history, available tools, and called workflows. The documentation estimates only ranges. Starting a fresh conversation for a new task and limiting tools can reduce context cost.
Bring Your Own Key changes model economics. The current guide says workflow AI nodes drop to one platform credit with BYOK, while Agent model credits receive a 50% reduction. Workflow base and tool costs still apply, and the model provider bills the customer's account separately.
At the review date, Gumloop pricing showed Pro at $37 per month with 7,400 base plus 12,600 bonus credits, displayed as 20,000 monthly credits. It included unlimited seats, five concurrent workflow runs, 25 concurrent Agent chats, connector policies, shared features, and one hosted MCP server. Enterprise added custom retention, RBAC, SAML or SCIM, audits, model controls, VPC options, and custom concurrency. The page offered a 14-day trial. A December 2025 official post described a 5,000-credit Free tier, but the live comparison no longer led with it; treat free availability as unconfirmed until the account screen shows it.
Credit overage is currently documented at $0.007 per credit, capped at twice the monthly allocation when enabled. Credits generally do not roll over outside Enterprise. Monitor loops and configure alerts before enabling automatic overage.
Privacy and training language
Gumloop marketing says customer data is not used to train AI models and describes zero-retention arrangements with external model providers. The privacy policy, effective June 26, 2026, is more specific: it says uploaded data, flows, and Agent chats from premium users are not used to train AI, and that OpenAI and Anthropic are contractually restricted from training on Gumloop API data.
Because the legal wording identifies premium users, a free or trial user should not assume identical treatment without confirmation. The policy also says personal data may be processed in the United States and retained as necessary for service, legal obligations, disputes, and agreements. Map logs, workflow data, Agent conversations, credentials, backups, provider requests, and downstream exports separately.
Governance and fit
Pro collaboration can make adoption efficient, but unlimited seats and shared credentials need organization rules. Define who may create connectors, share keys, expose MCP servers, enable overage, or run scraping. Enterprise controls matter when different teams should not inherit the same access.
Gumloop is a strong candidate for teams that need AI, web data, and visual workflows in one managed platform and can actively manage credits. Compare Activepieces for approval-first flows and an open-source path, Make for a broad scenario builder, and Bardeen for GTM-focused scraping and enrichment. See the AI automation and agent tools category for evaluation criteria.
Visit Gumloop
