Gumloop

Gumloop is a visual platform for building AI agents that work across business applications and company data. It combines multi-model AI, connectors, reusable skills, team collaboration and workflow orchestration without requiring every user to write code. Gumloop now leads with agents; its older flow-based automation features remain available but are labelled legacy on the current pricing page.

What is Gumloop?

Gumloop lets teams create agents that can understand company context, use connected tools and carry work from a request to an outcome. Users can assemble behaviour visually, connect SaaS applications, share credentials safely, choose among more than 35 models and expose agents through chats, integrations, MCP or other delivery surfaces.

The platform sits between a general AI assistant and a traditional automation builder. Agents handle interpretation and flexible reasoning, while workflows and connector rules provide more deterministic control. The strongest implementations use each approach where it fits instead of relying on an LLM for every decision.

Current Gumloop pricing

Gumloop currently publishes two plans:

• Pro: starts at $37 per month with a 14-day free trial, 20,000 included credits per month, unlimited agents, unlimited seats and teams, access to more than 35 models and five concurrent legacy workflow runs.
• Enterprise: custom pricing and credits, more deployment and administration controls, custom model-proxy support, organisation-wide connector guardrails, dedicated Slack support and optional embedded experts or Virtual Private Cloud deployment.

Pro supports 25 concurrent agent chats and one hosted MCP server on the current comparison page. Enterprise concurrency, MCP hosting and run volumes are negotiated.

Credits and the orchestration fee

Credits pay for the models and services used by an agent or workflow. The cost varies with the model, prompt size, output length and connected operation. Gumloop also lists an 8% orchestration fee on Pro, with discounts available for Enterprise.

This makes total cost more complex than the $37 headline. Calculate the full cost of representative tasks, including model consumption, retries, long contexts, data extraction and orchestration. Confirm how bring-your-own API keys affect credits and the orchestration fee, whether unused credits roll over and what happens when the monthly allocation is exceeded.

Key features

AI agents

Agents can reason over a request, draw on company knowledge and use connected services. Unlimited agent creation allows teams to build specialised assistants for sales, research, marketing, support or operations rather than overloading one universal bot.

Each agent should have a narrow purpose, approved tools and explicit boundaries. Limit write access, separate draft generation from final actions and require approval for customer communication, financial changes, deletion and publishing.

35+ models and custom proxies

Gumloop provides a broad model catalogue and supports bring-your-own API keys. Enterprise adds custom proxy support and model-access controls. This flexibility helps teams match speed, cost and capability to each task.

Provider changes can alter output quality and data handling. Maintain an approved-model list, record the model used for important runs and regression-test before switching a production agent.

Company Brain and connected data

Company Brain is designed to make organisational knowledge available to agents. The value depends on source quality, freshness and permissions. Connect only approved repositories, preserve access boundaries and make answers cite their sources when factual accuracy matters.

Test outdated, contradictory and restricted documents. An agent should acknowledge uncertainty rather than silently combining incompatible policies.

Skills and GitHub synchronisation

Reusable skills let teams standardise how agents perform tasks, while GitHub Skill Sync supports versioned development workflows. Treat skills like code: review changes, test edge cases, restrict distribution and assign ownership.

MCP, CLI and developer access

Gumloop supports MCP, hosted MCP servers, proxying and a command-line interface. These capabilities allow agents and external systems to use Gumloop tools programmatically. MCP servers can expose powerful actions, so authentication, tool descriptions, scopes and audit logging deserve the same attention as conventional APIs.

Legacy workflows

Existing visual flows remain useful for predictable multi-step automation, with queuing and concurrency controls. Gumloop now labels them legacy, so buyers should ask about the migration path, long-term support and whether a new agent should replace or call an existing flow.

Collaboration and administration

Both plans advertise unlimited seats and teams, which is unusual compared with per-user automation products. Shared credentials can reduce duplicated account connections, but administrators should ensure users cannot reveal or reuse underlying secrets outside approved agents.

Enterprise adds role-based access control, SCIM/SAML, an admin dashboard, audit logs, custom retention rules, AI spend insights, data exports, model access controls and organisation-wide connector policies. These controls are central for large deployments and should be tested during a pilot.

Security and privacy

Gumloop’s privacy policy states that premium-user uploads, flows and agent chats are not used to train AI models, and that its agreements with OpenAI and Anthropic prevent training on data sent through Gumloop’s API. Buyers using other providers or their own keys should confirm the relevant provider terms separately.

The company describes nightly database backups, weekly manual penetration testing, static analysis and Cloud Access Security Assessment scrutiny. Internal data access is restricted by default, sensitive accounts use two-factor authentication, and user credentials are isolated from other users.

Gumloop participates in the EU-U.S. Data Privacy Framework and UK Extension. Its policy explains that personal data can be processed in the United States, retained as necessary for stated purposes and shared with service providers. Users can request access or deletion subject to legal exceptions.

Enterprise buyers should request current security reports, subprocessors, incident-response commitments, encryption details and the scope of VPC deployment. A VPC option does not necessarily place every third-party model call inside the same network boundary.

Best use cases

• Lead generation, enrichment and qualification.
• CRM research and account preparation.
• Competitor monitoring and structured research.
• Content, SEO and ad-campaign operations.
• Call analysis and meeting preparation.
• Support agents connected to approved knowledge.
• Data analysis across spreadsheets and SaaS tools.

Advantages

• Unlimited agents, seats and teams on Pro.
• Broad model catalogue with bring-your-own-key support.
• Visual agent building plus reusable skills and connectors.
• MCP, CLI and developer-friendly integration options.
• Enterprise controls for identity, retention, spend and auditing.

Limitations

• Credit costs and the 8% orchestration fee complicate budgeting.
• The shift from workflows to agents creates migration questions.
• Agent reliability depends on model quality and tool design.
• Broad connector access can create large permission risks.
• Enterprise pricing and advanced controls require a sales process.
• Third-party model providers add separate data-governance considerations.

How to evaluate Gumloop

1. Use the 14-day trial for one real agent with measurable output.
2. Track credits, orchestration fees, provider cost, retries and human corrections.
3. Test normal, ambiguous, adversarial and permission-restricted requests.
4. Require approval for messages, records changes and destructive actions.
5. Review every connector scope and shared credential.
6. For Enterprise, test SSO, roles, audit logs, retention and model controls.
7. Ask for the roadmap for any legacy workflows you plan to keep.

Best alternatives

n8n offers deeper technical workflow control and self-hosting. Zapier is easier for broad SaaS automation. StackAI focuses on governed enterprise agents and deployment options. Microsoft Copilot Studio suits Microsoft-centred organisations. Relevance AI provides multi-agent workforce tools, while code frameworks such as LangGraph give engineering teams lower-level control.

Verdict

Gumloop is attractive for teams that want to build many connected AI agents without per-seat pricing. Its model choice, reusable skills, MCP support and enterprise controls make it more than a simple visual automation tool. The trade-off is variable credit economics and the operational risk of agents acting across business systems. Pilot one bounded use case, measure the total cost per approved outcome and establish connector, model and approval policies before expanding.