Gumloop

What is Gumloop?

Gumloop is a code-free platform for AI agents and visual workflow automation. Users can connect nodes for data input, web scraping, AI models, conditional logic, code, files, tables, and third-party applications on a canvas; they can also create agents that are capable of selecting tools autonomously, invoking various skills, and maintaining ongoing conversations.

It is suitable for creating deterministic workflows with clear inputs and outputs, as well as for handling open-ended tasks such as research, sales, customer service, operations, and data analysis. These two modes can also be combined: workflows can invoke Agents, and Agents can use pre-verified workflows as tools, thereby enabling both flexible reasoning and process control.

Visualizing Workbooks and Workflows

Gumloop uses a Workbook as its main working area. A Workbook can contain multiple tabs, nodes, and connections, with data being processed sequentially through the nodes.

Users can start with templates, or they can drag in input, output, AI, web page, data conversion, application action, and advanced control nodes to build automation from scratch.

The operation records keep track of the input, output, execution time, Credits, and error messages for each node, which facilitates the identification of faults. If a process fails, it can be resumed from the checkpoint, without having to redo all nodes from scratch every time.

When sharing, it is possible to set permissions for viewing, editing, and running, making it suitable for teams to work together on maintaining production processes.

Subflow modular reuse

Subflows allow a set of completed steps to be packaged into reusable sub-processes that can then be invoked by other workflows. For example, by encapsulating tasks such as retrieving company information, verifying domain names, summarizing business details, and entering data into CRM, the same logic can be reused across different sales processes.

Modularity helps to reduce the costs associated with extremely large canvases, as well as repeated configuration and maintenance tasks. When a Subflow is modified, all processes that reference it will adopt the updated logic; therefore, in a production environment it is necessary to first create a copy for testing and to establish fixed input/output structures, so as to prevent a single change from affecting multiple downstream processes.

AI Agents

Gumloop Agent is designed for tasks for which it is not possible to define all steps in advance. Users specify goals and constraints using natural language; they can also add connectors, skills, knowledge sources, sub-Agents, and workflow tools to the Agent. The Agent then decides on the next action based on the conversation and the context surrounding it.

The Agent supports continuous sessions, attachments, structured outputs, as well as reflection and instruction improvement. It is suitable for meeting preparation, lead research, request categorization, CRM management, data querying, and cross-application operations.

When sending emails, updating CRM systems, deleting records, or making public releases, confirmation steps and minimum permissions should still be in place.

How to choose an Agent and a workflow?

Field mapping, scheduled reports, file conversion, batch data extraction, and fixed approval workflows are more suitable for workflows, as each step in them is predictable and testable. Research and analysis, multiple rounds of questioning, tool selection, and unstructured requests are better suited for Agents.

Complex operations usually employ a hybrid approach: triggers initiate the workflow, agents handle the intermediate decision-making steps, and finally deterministic nodes are used for verification and data writing to the system.

Skills, knowledge sources, and evaluation

Agents can include reusable Skills, corporate knowledge sources, and specialized sub-Agents. Skills are used to document operational methods, knowledge sources provide business context, while sub-Agents handle specific areas of work.

Gumloop also provides Evaluations, which use sample questions, desired behaviors, and scoring results to continuously assess the performance of the Agent after its modifications.

Reflection will provide suggestions for improving instructions based on the results of the execution, but such automatic suggestions should not replace manual review. For high-risk operations, a fixed set of tests should be established to cover scenarios such as rejection of requests, permission issues, sensitive data, incorrect tool selection, and abnormal input.

AI models and built-in keys

The platform can call on more than 30 different models and suppliers, covering text processing, image analysis, as well as various options in terms of cost and speed. Users can select models based on specific nodes or Agent tasks, without being forced to use services from a single supplier for the entire process.

Plans Pro and above allow users to specify their own API Key from a model provider. According to the official documentation, using an internal key can reduce the cost associated with certain AI calls via Gumloop Credits by around 50%, but the model provider will still charge separately based on its own pricing.

When budgeting, it is necessary to take into account both the quota allocated by the Gumloop platform and the costs associated with external APIs.

Web scraping and data processing

Web-related nodes can capture page content, extract structured fields, process lists, and pass them to AI for classification or summarization; they are suitable for competitor monitoring, lead research, product information gathering, recruitment, content collection, and organizing public data. Loop Mode allows processing of each item in a list individually, while the Pro plan supports concurrent processing of up to 15 list items.

Changes in the website structure, login verification, measures to prevent automation, regional differences, and dynamic loading can all lead to failures. It is necessary to comply with the terms of service of the target website, as well as its rules regarding copyright, personal information, and the frequency of access, before using it.

For large-scale tasks, it is necessary to implement speed limits, retry mechanisms, duplicate removal, error protection, and manual sampling.

Connector and credential management

The official documentation lists over 150 different connectors that can be used to connect to services such as Gmail, Slack, Salesforce, HubSpot, Google Sheets, Notion, Airtable, and Google Drive. The Personal Connector is available only for individual use.

Team Connector can be shared among a team’s Agents and workflows.

Sharing connectors does not mean that all members should have the same permissions. Administrators should create dedicated service accounts for automated use, limit the scope of OAuth, rotate keys regularly, and revoke access promptly when members leave or a process is discontinued.

MCP capabilities

Gumloop allows the use of MCP tools in Agents, Agent Nodes, and custom MCP nodes, enabling external services or internal corporate tools to be integrated with Agents. The company also provides a Gumloop MCP Server, which allows compatible AI clients to discover and invoke the configured workflows and capabilities.

The Pro basic plan includes 1 hosting slot for an MCP Server, while the Enterprise version offers 5 such slots that can be expanded as needed. Before integrating a third-party MCP Server, it is necessary to examine its permissions, data flow, maintenance status, and the content it returns; unknown servers should not be connected to production environments using their credentials.

Triggers, Webhooks, and interfaces

Workflows can be executed manually, or they can be triggered by events such as scheduled times, Webhooks, emails, Slack messages, Google Drive files, Airtable, Google Sheets, or calendar entries. The interface allows workflows to be presented in the form of forms or team entry points, enabling those who are not responsible for setting them up to enter the necessary parameters and initiate the workflow.

Trigger Subscriptions allows interface users to save their own inputs and credentials, as well as to subscribe to scheduled or service-related events. The subscriber is responsible for covering the Credits associated with each automatic execution.

Timed workflows should be equipped with failure alerts, a limit on the number of executions, and idempotent writing mechanisms, in order to prevent duplicate emails or duplicate CRM records from being created as a result of repeated attempts.

Hosting Agent page and team entry point

Pro offers a managed Agent page, custom Slack applications, Agent email addresses, Artifacts, image generation capabilities, a code sandbox, and a Secrets Vault. Teams can use these Agents as entry points for internal conversations, or have them receive requests via Slack and email, thereby avoiding the need for each team member to edit the underlying workflows directly.

The managed entry point simplifies distribution, but it does not automatically handle permission management. Before making it available to external users, it is necessary to restrict the tools that can be used, the types of files that can be uploaded, session duration, sensitive terms, and the frequency of calls; furthermore, it is important to ensure that the generated files do not reveal any internal data.

API and CLI

Public APIs can initiate workflows or Agents asynchronously, continue existing Agent sessions, poll for execution results, and retrieve execution history. Once an Agent is started, an interaction_id is returned; the caller can then use the status interface to obtain the final results.

Authentication can be done using a personal API Key or OAuth.

The official CLI is used to manage and invoke Gumloop’s capabilities from the command line, and it is suitable for development, deployment, and automated testing. API calls also consume Credits associated with the respective processes and Agents.

Production access should securely store keys, handle verification callbacks, and manage queued as well as failed states.

How are Credits calculated?

Credits are Gumloop’s unified unit of measurement; currently, 1 dollar allows for the purchase of 200 Credits, which means each Credit is worth approximately 0.005 dollars. Regular nodes may be available free of charge or at a price specified per node, while agent chats are billed based on a combination of costs related to the model, tools, processing time, and orchestration.

For the Agent’s Chat and Reasoning functions, Credits are calculated based on the actual cost of the model, rounded to the nearest whole number; at least 1 Credit is charged for each successful tool call, in addition to the cost associated with that tool itself, while no charge is applied for failed tool calls.

The cost for active sessions is calculated at 5 Credits per minute, with a minimum of 1 Credit per response; the orchestration fee amounts to 8% of the total cost associated with chat reasoning, computation, and tool calls.

When an Agent invokes a complete workflow, charges are applied based on the cost of each node within that workflow; no additional fees for the Agent’s computing and orchestration services are added.

Pro allows for exceeding the allocated quota; any usage beyond that is charged at 0.005 dollars per Credit, with a maximum limit of twice the planned quota. When excess usage is not enabled, operations that require Credits will stop once the quota is exhausted.

The unused monthly credit is usually not carried over; Enterprise clients can negotiate the rules for carrying it over.

Free account

According to the official details regarding price adjustments for 2026, the Free account provides 5,000 Credits per month, which is an increase from the 2,000 Credits offered in the previous plan. The free tier is suitable for using a drawing canvas, creating prototypes, and running simple processes; however, it offers fewer capabilities in terms of concurrency, team management, and advanced functions compared to the Pro version.

Some older help pages may still mention options such as Free 2,000 Credits, Solo, or Team; this is due to differences in the timing of document updates. To determine the current benefits, one should refer to the account’s Subscription page and the latest Pricing page.

Pro price and benefits

The starting price for Pro is $37 per month; the basic plan offers 7,400 base Credits plus 12,600 bonus Credits, for a total of 20,000 Credits per month.

Pro offers no limits on the number of seats; it includes 5 concurrent workflows, 25 simultaneous Agent conversations, an unlimited number of Teams, a unified billing system, unlimited workflows, Agents, Skills, Knowledge Sources, Artifacts, Evaluations, triggers, Interfaces, Webhooks, as well as various capabilities for team distribution.

Pro can be upgraded to a higher tier using the quota slider; according to the official documentation, the amount available can range from 20,000 per month up to 1.5 million Credits. When upgrading midway, the new quota is added to the existing balance rather than resetting it to zero.

The specific discounts, annual payments, reward Credits, and taxes are subject to the settlement page.

Enterprise plan

Enterprise offers customized pricing and includes features such as RBAC, SCIM and SAML, an administrator panel, usage analysis, AI-driven insights into spending, audit logs, custom data retention options, security reports, data export capabilities, incognito mode, model access control, VPC, workflow queuing, custom concurrency settings, and dedicated support.

Before making a purchase, a company should verify the credit transfers, concurrent processing capacity, MCP hosting volume, VPC regions, data retention policies, zero retention for models, as well as the response times and service levels offered. The capabilities listed in the price table do not mean that all of them are enabled by default under a contract.

GitHub and the open-source status

Gumloop has an official GitHub organization; the public repositories there contain projects such as gumloop-mcp and gumloop-celery, some of which are forks of upstream open-source projects. These repositories do not include the complete source code for Gumloop’s visualization canvas, the Agent runtime, the connector platform, and the cloud services.

Therefore, Gumloop’s core products should be labeled as hosted SaaS solutions that are not open source; it cannot be claimed that they can be deployed in a fully private environment just because there is an official GitHub account. If operation within an internal network is required, it is necessary to consult the Enterprise sales team regarding VPC and data boundaries.

Gumloop usage guide

Complete a basic task.

1. Define the code tasks to be completed in Gumloop, the scope of the repository, and the acceptance criteria.
2. Connect to or import the test project, and first back up the current branch;
3. First, use visual Workbooks and workflows to create a plan, and then verify the modified files.
4. Use the Subflow module for modular reuse to make minor changes;
5. Run tests, static checks, and builds; unverified code is not accepted directly.
6. Manually check permissions, keys, dependencies, and handle exceptions before merging;

Create reusable professional workflows

1. Select a low-risk, real-world project as a template;
2. Fix the order of using visual Workbooks and workflows, modular reuse of Subflows, and AI agents.
3. Record the environment, model, prompts, and failure conditions;
4. Set up manual approval for write, deploy, and delete actions;
5. Compare speed, cost, test pass rate, and the amount of rework;
6. Expand to a team or production environment only after stability has been verified;

Which users is it suitable for?

• We hope to use visual canvases to help the operations, marketing, and data teams manage AI workflows;
• Sales teams that need clue-based research, CRM updates, and automation for emails and Slack;
• Researchers who need to scrape public web pages and process them structurally;
• Companies that wish to handle open requests through agents while maintaining deterministic processes;
• Developers who need API, Webhook, CLI, and MCP to connect to internal systems.

Product advantages

• Workflows and adaptive Agents can be combined in a two-way manner;
• The code-free canvas, Subflow, checkpoints, and execution history are suitable for maintaining complex processes;
• There are a large number of connectors, and distinctions are made between individual and team credentials;
• Supports MCP, API, CLI, Webhook, and various triggers;
• Pro offers unlimited seats, ideal for teams to share workflows and Credits;
• It provides capabilities for assessment, reflection, authorization, and corporate governance.

Restrictions and Precautions

• Credits are accumulated based on nodes, models, calculation time for Agents, as well as costs related to tools and orchestration; long conversations, large-scale data extraction, and third-party data nodes with high pricing can lead to a rapid consumption of these credits.
• The reward Credits, price slider, and package benefits can change, so it is necessary to check the account settlement page before making a purchase.
• AI may choose the wrong tools, generate incorrect fields, or perform actions that do not meet the expected outcomes;
• Web automation can be affected by changes to the target website’s design.
• Connectors and MCPs expand the scope of data access;
• When it comes to payments, deletions, data distribution, customer information, and production databases, approval processes, minimal permissions, test environments, audit logs, and a rollback mechanism should be employed.

Frequently Asked Questions

Is Gumloop free?

There is a free account; the latest official specification is 5,000 Credits per month. The 2,000 Credits mentioned on the old page refer to an older scheme – the actual amount is determined by the subscription details on the account page.

How much is Gumloop?

The Pro basic plan costs $37 per month and includes 20,000 Credits as well as unlimited seats; higher credit packages can be selected by the user themselves.

Enterprise comes with a customized price.

Are the Solo and Team versions of Gumloop still available?

The current public pricing has combined the Solo and Team versions into a Pro version. The old prices of $37 for the Solo and $244 for the Team, which are commonly seen online, no longer reflect the current pricing for new purchases.

Does Gumloop support APIs and MCP?

Supported. The public API can initiate workflows and Agents, as well as continue sessions and retrieve query results;

It also provides an MCP Server, access to MCP tools, and a CLI.

Is Gumloop open source?

The core platform is not open source. The official GitHub site provides a small number of projects and forks, but these do not represent the complete source code of the Gumloop product.