CtrlK
Gumloop is a no-code AI workflow automation platform that lets teams build, deploy, and manage AI agents and automated workflows through a visual drag-and-drop interface. Founded in April 2023 by Max Brodeur-Urbas and Rahul Behal, two McGill University classmates who previously worked at Microsoft and Amazon respectively, the company started as a side project in a Vancouver bedroom before entering Y Combinator's Winter 2024 (YC W24) batch. Gumloop has since raised $70.6 million in total funding, relocated its headquarters to San Francisco, and counts Shopify, Ramp, Gusto, Instacart, and Opendoor among its enterprise customers.
The platform positions itself as a middle ground between consumer-grade automation tools like Zapier and developer-oriented systems like n8n: capable enough for complex AI-driven workflows, but accessible enough that non-engineers can build production pipelines without writing code. Its product line includes a visual workflow builder, an AI agent framework, a Chrome extension for browser automation, and Gumstack, an enterprise security and observability layer that monitors AI tool usage across an organization.
Max Brodeur-Urbas and Rahul Behal met as students at McGill University in Montreal. After graduating, both moved into software engineering roles at large tech companies: Brodeur-Urbas joined Microsoft, and Behal went to Amazon Web Services as a machine learning engineer. They reconnected at an AI meetup in Vancouver in 2023 and began building together as a side project.
The company launched in April 2023 under the name AgentHub. The original product was a UI wrapper around AutoGPT, the early open-source autonomous agent framework that attracted significant attention in early 2023. The founders built it primarily to serve a non-technical Discord community that wanted to experiment with AI automation but found AutoGPT too complex to configure directly.
As the product matured, the founders moved away from the autonomous agent model. They concluded that fully autonomous AI workflows were too unreliable for production use and pivoted toward a more structured approach: a visual node-based canvas where users define the logic explicitly and AI handles specific tasks within those boundaries. This philosophical shift is captured in a line Brodeur-Urbas has repeated in interviews: "Leaving specific workflows completely up to AI is not realistic."
In May 2024, the company renamed itself from AgentHub to Gumloop. The founders cited several practical reasons. The name AgentHub was often misheard when spoken quickly, and the .dev domain signaled a developer-only product that contradicted the no-code positioning. More substantively, users who knew other agent frameworks arrived expecting autonomous bots and were confused by what they found. The word "agent" carried baggage the product didn't match.
The name Gumloop was chosen partly on the advice of Paul Graham's essay on startup naming. "Gum" is meant to evoke the sticky connective tissue that binds integrations and data together; "loop" refers to the platform's core function of running automations repeatedly at scale. The founders also noted that the lowercase "gumloop" has a visual symmetry (3-3 letter structure) that works well in branding, and that the .com and .ai domains were both available.
The company participated in Y Combinator's Winter 2024 batch as AgentHub and appeared at the W24 Demo Day under that name before the rebrand. Gustaf Alstromer served as the primary YC partner. Gumloop was one of a relatively small number of Canadian companies in that cohort.
Gumloop has raised $70.6 million across four rounds.
In July 2024, shortly after the rebrand, Gumloop closed a $3.1 million seed round led by First Round Capital. Y Combinator also participated, along with a set of angel investors: Max Mullen (co-founder of Instacart), Arash Ferdowsi (co-founder of Dropbox), and Richard Aberman (co-founder of WePay). At the time the seed closed, the team consisted of the two co-founders and one intern. The company was executing nearly half a million nodes per day across its user base, and one early customer was saving over $200,000 annually through workflows built on the platform.
In January 2025, Gumloop raised $17 million in a Series A round led by Nexus Venture Partners, which took a board seat. First Round Capital and Y Combinator returned. Additional angels included Bryant Chou (co-founder of Webflow), Reynold Xin (co-founder of Databricks), and Shaan Puri (creator of the My First Million podcast). The Canadian dollar equivalent was approximately $24.5 million CAD, which drew significant coverage in the Canadian tech press through BetaKit.
Around the time of the Series A, Brodeur-Urbas publicly articulated an unusual ambition: building a $1 billion company with a maximum of ten employees. He pointed to San Francisco's talent density as the reason for relocating from Vancouver, where the team would grow from three people (Brodeur-Urbas, Behal, and one hire) to a small but senior group.
In March 2026, Gumloop raised $50 million in a Series B led by Benchmark. Everett Randle, a general partner at Benchmark, led the deal; it was his first investment at the firm. Returning investors included Nexus Venture Partners, First Round Capital, Y Combinator, BoxGroup, and The Cannon Project. Shopify Ventures joined the round as a new strategic investor, reflecting Shopify's position as both an investor and a customer.
By the time of the Series B, Gumloop had grown to 25 employees. The company stated it would use the capital to build out a dedicated sales organization for enterprise customers and expand its engineering team to deepen platform capabilities and add integrations.
Gumloop's platform has three main product areas: workflows (structured visual automations), agents (AI systems that reason over available tools), and Gumstack (enterprise security and observability). The underlying interface is a shared canvas where all three converge.
The central interface is a visual node-based canvas. Users drag nodes onto the canvas from a sidebar, then draw connections between them to define how data flows from one step to the next. Each node represents a discrete operation: fetching a URL, reading a PDF, running an LLM prompt, parsing JSON, sending an email, or querying a database. Connecting a web scraper node to an LLM node to a Google Sheets node creates a workflow that scrapes a site, summarizes what it found, and logs the result to a spreadsheet.
Gumloop describes this as "backend coding by dragging visually appealing blocks onto a canvas." The canvas supports loops (running a node over a list of inputs), conditionals (routing data down different branches based on logic), and subflows (nested workflows that function like reusable functions in code). The platform executes these in the cloud, so workflows run without a browser or local machine staying open.
Gumloop ships over 100 pre-built nodes organized into categories.
AI nodes include: LLM calls (supporting OpenAI, Anthropic, Google Gemini, and open-source models), AI text extraction, AI data classification, AI summarization, image generation, and AI web research. Users can route different tasks to different models within the same workflow, for example using a cheaper model for classification and a more capable one for drafting.
Data nodes cover: web scraping, PDF reading (including scanned documents via AI vision), CSV parsing, JSON transformation, and database queries. The PDF reader handles both native text PDFs and image-based documents.
Integration nodes connect to external services directly. The native integration library covers Google Sheets, Gmail, Google Calendar, Google Docs, Slack, GitHub, Jira, Salesforce, YouTube, and others. Gumloop also supports webhooks for triggering workflows from external systems and REST API calls for services without dedicated nodes.
In May 2025, Gumloop introduced MCP (Model Context Protocol) nodes, which represent a significant architectural shift. Rather than requiring Gumloop to hand-code each integration, MCP nodes let users describe in natural language what they want from a service, and the system dynamically generates the Python code to handle the API call. The initial MCP rollout included Slack, Gmail, Google Calendar, Google Docs, Jira, Salesforce, Reddit, and GitHub, with over 70 additional integrations queued. MCP adoption means Gumloop can expand its integration coverage at a much faster rate than was possible with the older node-by-node approach.
In 2025, Gumloop added an agent layer on top of the workflow canvas. Where workflows execute a fixed sequence of nodes, agents use LLM reasoning to decide which tools to use and in what order. A user defines an agent's goal in a system prompt, connects tools (MCP servers, Gumloop workflows, or both), and the agent figures out how to combine them.
Agents can be deployed directly to Slack, Microsoft Teams, or email, where they respond to messages from colleagues. In Slack, agents operate in shared channels, which means teammates can see what the agent does and learn from how it solves problems. Gumloop frames this visibility as a feature: knowledge about what agents can do spreads organically through teams rather than being siloed with whoever built the agent.
The platform distinguishes between workflows and agents by use case: workflows handle "frequent, repeatable tasks at scale" with predictable outputs; agents handle open-ended, complex tasks that require judgment. Gumloop's architecture lets agents use workflows as tools, combining both approaches in a single system.
Skills are reusable capabilities that can be attached to agents. A skill might encode a particular scoring logic, a document template, or a standard API call pattern. Once created, a skill can be attached to multiple agents, avoiding the need to rebuild the same logic repeatedly. SiliconAngle described skills as "customizable extensions that enhance agent functionality" and gave the example of a severity-ranking skill for support ticket triage.
Gumloop ships a Chrome extension that allows workflows to interact with the browser. Users can upload the content of browser tabs directly into a workflow for processing, and can schedule the browser to visit pages at set intervals and extract data. This enables automation patterns that require a logged-in browser session, such as pulling data from internal dashboards or web applications that do not expose APIs.
Gumstack is Gumloop's enterprise security and observability product, positioned as a separate layer from the automation platform itself. The problem Gumstack addresses is that large organizations now have employees using dozens of AI tools simultaneously (Claude Code, ChatGPT, Cursor, internal agents) with no centralized visibility into what data those tools are accessing or what actions they are taking.
Gumstack provides a single dashboard where administrators can:
Gumstack supports SSO and SCIM for identity management integration. It is compatible with third-party AI tools beyond Gumloop itself: Gumstack can monitor tool calls from Claude Code, ChatGPT, and Cursor, not just from Gumloop agents. The company has described Gumstack as the product of two years of work on authentication and observability infrastructure.
CEO Brodeur-Urbas said in March 2026 that recent LLM capability improvements "accelerated our roadmap" by making the agents powerful enough to justify the security layer Gumloop had already built.
Gumloop supports over 400 integrations through a combination of native nodes, MCP servers, and webhook-based connections. The major categories are:
| Category | Services |
|---|---|
| Communication | Slack, Microsoft Teams, Gmail, Outlook, email SMTP |
| Productivity | Google Docs, Google Sheets, Google Calendar, Google Slides, Google Tasks, Notion, Jira, Airtable |
| Developer tools | GitHub, Google BigQuery, webhooks, REST API |
| CRM and sales | Salesforce, HubSpot, Apollo |
| Content and media | YouTube, Reddit, Google Search, web scraping |
| Storage | Google Drive, Dropbox |
| AI models | OpenAI, Anthropic Claude, Google Gemini, open-source models |
The MCP node system, introduced in May 2025, means new integrations can be added by the community or by connecting any MCP-compliant server without waiting for Gumloop to ship a dedicated node. Gumloop also exposes its own API as an MCP server, so AI models and coding assistants can trigger and manage Gumloop workflows programmatically.
Gumloop is frequently used to automate multi-source research. A common pattern is a workflow that takes a list of company names, visits each company's website to extract product descriptions, team size, and recent news, runs the extracted text through an LLM for summarization or scoring, and outputs a structured report to Google Sheets or Notion. More complex versions add LinkedIn scraping, Google Maps lookups, or calls to data enrichment services like Apollo.
One customer example from Gumloop's seed announcement involved an accounting firm that built a workflow to identify PDFs in email inboxes, extract invoice data, and push it to a spreadsheet. This use case eliminated a significant amount of manual data entry.
Sales and marketing teams use Gumloop to enrich lead lists before outreach. A workflow might take a raw list of signups or webinar attendees, visit each person's LinkedIn profile or company website, extract relevant signals (job title, company size, industry, recent funding), score the lead using an LLM, and generate a personalized outreach email. Gumloop's own blog describes enriching 100 contacts in minutes rather than hours.
Local business prospecting is another sales application: users build workflows that pull business listings from Google Maps using a keyword and location, enrich each business with contact info via Apollo, and export everything to a spreadsheet ready for outreach.
Content and SEO teams use Gumloop for research-heavy content workflows. An SEO agent can pull keyword rankings, scrape top-ranking competitor pages, extract their structure and key arguments, and produce a content brief that incorporates those insights. Users have built workflows that check keyword positions every morning and send a Slack digest. More elaborate workflows pull from a company's past content archive and brand guidelines, generate a draft, verify factual claims against source URLs, and run the result through a brand voice check before sending it for human review.
Gumloop's blog describes one user who automated an entire YouTube channel workflow, from generating video ideas based on trending topics to scripting, thumbnail generation, and scheduling.
Enterprises that handle large volumes of documents, including law firms, logistics companies, and organizations that interact with government forms, use Gumloop to extract structured data from PDFs, contracts, and forms. The platform's PDF reader handles both native text documents and scanned image-based PDFs using AI vision. Extracted data flows into downstream systems automatically, removing manual keying. Gumloop's document workflow automation page specifically calls out invoice processing, contract review, and form extraction as target applications.
Customers also use Gumloop for internal operations tasks: ticket categorization and routing, hiring automation (screening resumes, scheduling interviews), and generating reports from internal data sources. Agents deployed to Slack can handle employee queries autonomously, for example looking up HR policy documents or pulling data from internal dashboards.
Gumloop uses a credit-based pricing model with three tiers.
| Plan | Price | Credits per month | Seats | Concurrent runs | Notable features |
|---|---|---|---|---|---|
| Free | $0 | 5,000 | 1 | 2 | Unlimited workflows and agents, forum support |
| Pro | $37/month | 20,000+ | Unlimited | 5 | Unlimited teams, MCP Server hosting (1), MCP proxying (3), app policies and guardrails, team usage analytics |
| Enterprise | Custom | Custom | Custom | Custom | SAML/SCIM, RBAC, audit logs, VPC, custom data retention, AI model access controls, Gumstack features |
Annual billing reduces Pro plan cost by 20%. Credits are the unit of consumption: each node execution consumes credits, with the cost varying by node type. LLM calls consume significantly more credits than simple data transformation steps. Running a node across a list of items multiplies the credit cost by the list length, which can surprise users who do not account for batch processing in their credit budget. Credit overages on the Pro plan are billed at an additional per-credit rate.
The Free plan is intended for exploration and light workflows. Pro covers most individual users and small teams with moderate automation needs. Enterprise adds the governance and compliance features (SSO, audit logs, Gumstack integration) that large organizations require.
Gumloop competes primarily with n8n, Zapier, Lindy, and Make (formerly Integromat) in the workflow automation space. The competitive landscape breaks roughly along two axes: how much technical knowledge is required, and how deeply AI is integrated into the automation logic.
| Feature | Gumloop | Zapier | n8n | Lindy |
|---|---|---|---|---|
| Launch year | 2023 | 2011 | 2019 | 2023 |
| Hosting | Cloud only | Cloud only | Self-hosted and cloud | Cloud only |
| Integrations | 400+ | 6,000+ | 800+ | 200+ |
| AI integration | Native, central | Peripheral (plugin step) | Plugin support | Native, central |
| Self-hosting | No | No | Yes | No |
| Workflow logic | Visual canvas, branches, loops | Linear triggers and actions | Complex branching, loops | Conversation-based agents |
| Free tier | Yes (5,000 credits/month) | Yes (100 tasks/month) | Yes (self-hosted) | Yes (limited) |
| Pro pricing | $37/month | $19.99-$799/month | ~$20/month (cloud) | $49.99/month |
| Target user | Growth and ops teams, SMBs | Non-technical SMBs | Technical teams, developers | Non-technical individuals |
| Enterprise features | Gumstack, SAML, VPC | SSO, advanced logs | LDAP, audit logs | SSO |
Zapier has the largest integration library by a wide margin (6,000+ vs. Gumloop's 400+) and a decade of reliability track record. For organizations that need simple, well-tested automations across many SaaS tools, Zapier's breadth is difficult to match. The tradeoff is cost (Zapier pricing scales steeply with task volume), limited support for non-linear logic, and AI that is bolted on as a step rather than woven into the design.
n8n is the strongest choice for technical teams that want full control. Its open-source self-hosted version gives organizations complete data sovereignty, and its workflow engine handles complex branching logic that neither Gumloop nor Zapier can match cleanly. The downside is that n8n requires DevOps investment to maintain and the learning curve is steep for non-engineers.
Lindy is closer to Gumloop in its AI-native orientation. Lindy leans more toward conversational agent building (describing workflows in natural language rather than drawing them on a canvas) and is generally positioned for individual knowledge workers rather than ops teams building shared infrastructure.
Gumloop's differentiated position is the combination of a visual canvas (which makes complex logic inspectable and debuggable), native AI nodes that can be used anywhere in a workflow, and the Gumstack observability layer for enterprise governance. The canvas approach means workflows are easier to audit than agent systems that operate as black boxes.
Gumloop's disclosed enterprise customers include Shopify, Ramp, Gusto, Samsara, Instacart, Opendoor, and Rippling. Shopify is notable as both a customer and an investor (via Shopify Ventures), having participated in the Series B.
The customer base spans multiple use cases. Ramp (corporate card and expense management) and Gusto (payroll and HR software) are likely using Gumloop for internal operations automation. Samsara (fleet and operations management) and Opendoor (real estate) suggest use cases involving document processing and data workflows. Instacart's involvement is also reflected in angel investor Max Mullen's participation in multiple funding rounds.
Several limitations come up consistently in user reviews and independent assessments.
The credit system can be difficult to budget in advance. Credits are consumed per node execution, and batch processing multiplies costs in ways that are not always obvious from the pricing page. One review calculated that enriching 100 contacts at 60 credits each costs 6,001 credits, nearly a third of a Pro plan's monthly allowance. Users who run large lists frequently find the Pro plan ceiling constraining and face steep overage charges.
The integration library, while growing rapidly through MCP, is still much smaller than Zapier's 6,000+ connectors. Organizations with heavily customized SaaS stacks may find gaps.
Complex workflows can run slowly. User reviews note that long, multi-step workflows experience latency, particularly when running many LLM calls in sequence.
The platform has no live chat support. Free and Pro users interact with a community forum or wait for email responses. Only Enterprise customers get dedicated support.
Gumloop does not support self-hosting. Organizations with strict data residency requirements can use the Enterprise plan's Virtual Private Cloud option, but there is no on-premises deployment path equivalent to n8n's self-hosted version.
The no-code promise has limits in practice. User reviews suggest that while non-technical users can build simple automations quickly, complex production workflows with error handling, conditional routing, and data transformation require a meaningful time investment. Some users report needing 50-100 hours of practice before feeling confident with the platform.