# Gumloop Build and share work agents: connect apps, perform browser tasks and coordinate recurring workflows. Canonical profile: https://ki-agentenkompass.de/en/agents/gumloop/ Editor: [Michael Meierhoff](https://ki-agentenkompass.de/en/about/) Sources reviewed: 2026-10-12 Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests. ## Overview Gumloop is a platform for building and jointly operating agents. - Provider: AgentHub Inc. - Product family: Gumloop - Profile type: Platform for work agents - Environment: Managed cloud and connected work tools - Connections: App connectors, browser, Slack, Microsoft Teams and Gmail - Setup: Create an account, configure an agent and connect the required applications. - Pricing model: Pro from USD 37/month with 20,000 credits; an 8% orchestration fee is listed. ## Key features ### Build agents Teams build agents for tasks across connected business applications. Source: https://www.gumloop.com/ ### Browser actions Agents can interact with websites and hand control to a person when needed. Source: https://www.gumloop.com/ ### Events and schedules App events and recurring tasks can trigger agents. Source: https://www.gumloop.com/ ### Share with your team Agents can be shared with different usage and editing permissions. Source: https://www.gumloop.com/ ### Track usage The platform shows usage and spending by agent and model. Source: https://www.gumloop.com/ ## Example workflows ### Prepare inquiries - Your input: A connected CRM and rules for new inquiries. - Possible result: A structured processing suggestion for new inquiries. - What to check: Check customer data, action permissions and external messages. ### Create a weekly overview - Your input: Approved data sources and a schedule. - Possible result: A recurring overview from connected applications. - What to check: Check sources, schedule and usage. ## Requirements and limits ### Authorize connections Requirement: Access to apps and browser accounts must be configured. Source: https://www.gumloop.com/ ### Account for consumption Limitation: Credits and usage-based costs limit what is included in the base price. Source: https://www.gumloop.com/pricing ### Enterprise terms Not yet confirmed: Enterprise pricing and optional services are agreed individually. Source: https://www.gumloop.com/pricing ## A good fit if … - you want to start agents by cron, webhook or Gmail, Slack and Teams events - you want to set per connector whether writes and deletes need confirmation - you want to run agents as a team with shared or personal accounts and a shared workspace ## Less suitable if … - you need SSO, audit logging or connector policies without an Enterprise plan - you need statements on data region or hosting in Europe; the reviewed pages say nothing on this Editorial assessment based on provider documentation. Not a product test. ## Usable AI models Fixed choice across providers You pick a model or preset from the list Gumloop maintains, or leave it to the Auto router. Own keys work from the Pro plan for OpenAI, Anthropic, Google, Perplexity, xAI and Fireworks; the docs describe no way to add arbitrary models. On Enterprise, admins can restrict models and route supported providers through their own proxy. - Model families named: Anthropic (Claude), OpenAI (GPT), Google (Gemini), Meta (Llama, Muse), xAI (Grok), Moonshot (Kimi), DeepSeek, MiniMax, Z.ai (GLM) - Model choice: Automatic or your choice - Your own API key: Yes - Router or custom endpoint: No - Local models: Not stated - Depends on: Plan, Admin approval Documented · Sources accessed on: 2026-11-05 - [Gumloop: AI models](https://docs.gumloop.com/core-concepts/ai_models) - [Gumloop: Use Your Own LLM Key](https://docs.gumloop.com/help/connectors/use-your-own-llm-key) ## Europe & regional availability Official product and regional sources reviewed. The stated open points still require confirmation from the provider. Reviewed scope: Product access, feature limits and regional evidence Reviewed on: 2026-10-19 ### Europe: specific availability unconfirmed Availability still unclear Gumloop describes agents and addresses EU/EEA users in its privacy policy, which states that data is transferred to the US. Complete country and feature availability for Europe remains unconfirmed. Region covered: EU/EEA, United Kingdom and Switzerland Documented reason: The reviewed sources do not provide a complete list of supported countries for this. Official source: https://www.gumloop.com/ Official source: https://www.gumloop.com/privacy-policy ## Accounts & permissions Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use. Configuration covered: Gumloop agents in the managed cloud, individual, team or Enterprise organisation Sources reviewed: 2026-11-05 ### Account used Documented: By default an agent uses the default account of the person running it. The team's shared default account or specific pinned accounts can be used instead. Triggers run with the credentials of the person who created them. With agent-owned credentials, everyone with access to the agent uses the pinned accounts. Evidence: https://docs.gumloop.com/core-concepts/credentials.md Evidence: https://docs.gumloop.com/core-concepts/agent_triggers ### Admin approval Documented with conditions: Connector policies, custom roles for connectors, tools and models, audit logging, and SSO with SCIM are listed as Enterprise features. Agent-owned credentials are restricted by default for organisations; an admin enables them per custom role. Evidence: https://docs.gumloop.com/enterprise-features/overview.md Evidence: https://docs.gumloop.com/core-concepts/credentials.md ### Data access & team sharing Documented: Personal connectors are private; team connectors are scoped to one team and its members. Files in the team workspace are visible to all other members, while the personal workspace is never visible to others. Evidence: https://docs.gumloop.com/core-concepts/credentials.md Evidence: https://docs.gumloop.com/core-concepts/agent_sandbox_and_secrets.md ## Prices and included usage Pricing checked: 2026-10-12 Pro from USD 37/month with 20,000 credits; an 8% orchestration fee is listed. ### Pro From USD 37/month 20,000 credits per month; 14-day trial. Price source: https://www.gumloop.com/pricing ### Enterprise Custom Custom allowances and enterprise features. Price source: https://www.gumloop.com/pricing The provider describes tokens and compute at cost plus an 8% orchestration fee; total costs depend on usage. Pricing details: https://www.gumloop.com/pricing ## Operation & oversight Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured. Deployment covered: Hosted agents in the Gumloop cloud with triggers, schedules and webhooks Sources reviewed: 2026-11-05 ### Files and working state Documented: Each conversation runs in its own cloud sandbox, and new conversations start with a fresh working directory. Conversation files are saved with the conversation; team and personal workspaces and installed packages persist across conversations. Python variables are not archived. Evidence: https://docs.gumloop.com/core-concepts/agent_sandbox_and_secrets.md ### Sign-ins Documented: OAuth tokens renew automatically while the saved authorisation is still valid. An account that cannot work as it is shows a "Reauth needed" badge; you authorise it again and retry the task. For temporary failures the docs advise simply retrying first. Evidence: https://docs.gumloop.com/core-concepts/credentials.md ### Starting further runs Documented: Runs start on a cron schedule or once, through a webhook URL, or on app events such as Gmail, Outlook, Slack and Teams. With AI Managed on, the agent can create and manage its own triggers; webhook changes need your approval. Evidence: https://docs.gumloop.com/core-concepts/agent_triggers ### Human intervention Documented: Per connector you set whether tools are always allowed, need confirmation each time, or only for writes and deletes. Approvals appear in the chat, the notification inbox, the tasks page and the Slack thread. A gated call does not run without explicit consent. Evidence: https://docs.gumloop.com/core-concepts/human_in_the_loop.md ### Detecting errors Documented: A trigger's task history lists all its runs and filters by completed, approval required and failed. After three consecutive failed runs a trigger is deactivated, and the owner is told by email, Slack DM and in the app. Evidence: https://docs.gumloop.com/core-concepts/agent_triggers ### Resuming after failures Documented: An agent paused for approval saves its state and resumes after your decision. Deactivated triggers can be re-enabled at any time. The approvals page does not describe automatic retries of failed runs. Evidence: https://docs.gumloop.com/core-concepts/human_in_the_loop.md Evidence: https://docs.gumloop.com/core-concepts/agent_triggers ### Editorial assessment Failures and approvals are surfaced, and triggers switch themselves off after repeated errors. Oversight mainly means re-authorising connections when needed and re-enabling deactivated triggers. ## Memory and conversation continuity Scope: Agent reflections and separate Brain RAG Sources accessed: 2026-11-04 Reflections improve saved agent configuration; Brain searches company knowledge. ### Storage and scope Partly documented: Durable changes affect saved agent instructions, skills and triggers; their backend and a personal fact store are not established here. - [Reflections](https://docs.gumloop.com/core-concepts/reflections) ### Selection and timing Documented: Reflections periodically inspect task history and configuration and propose changes; automatic application is configurable. - [Reflections](https://docs.gumloop.com/core-concepts/reflections) ### Inspect, edit and delete Partly documented: Suggestions and rationales can be reviewed; changes can be applied and reverted. Individual personal memory facts are not described. - [Reflections](https://docs.gumloop.com/core-concepts/reflections) ### Agent and channel boundaries Documented: Reflections belong to the respective agent; Brain accesses shared company knowledge. - [Reflections](https://docs.gumloop.com/core-concepts/reflections) - [Agents and Brain](https://docs.gumloop.com/core-concepts/agents) ### Retrieval Documented: Saved instructions and skills affect later work; Brain separately retrieves documents and messages with citations. - [Reflections](https://docs.gumloop.com/core-concepts/reflections) - [Agents and Brain](https://docs.gumloop.com/core-concepts/agents) ### Session reset and compaction Partly documented: Configuration changes affect later tasks. Conversation-context reset and compaction are not specified in these sources. - [Reflections](https://docs.gumloop.com/core-concepts/reflections) - [Agents and Brain](https://docs.gumloop.com/core-concepts/agents) ## Context Access to apps and browser accounts must be configured. Official product documentation reviewed. Workflows and assessments are editorial; hands-on product tests are outside the scope of this directory. ## Open points Europe, availability still unclear: - Europe: specific availability unconfirmed Requirements and limits, not yet confirmed: - Enterprise terms ## Sources - [Gumloop: Build AI agents for work](https://www.gumloop.com/) - [A simple, transparent pricing model](https://gusto.gumloop.com/blog/transparent-pricing) - [Pricing | Gumloop](https://www.gumloop.com/pricing) ## Provider Company behind the Gumloop platform. - Operator: AgentHub Inc. - Organization: Company - Website: https://www.gumloop.com/ - Reviewed: 2026-10-12 - Provider source: https://www.gumloop.com/ [Sources and method](https://ki-agentenkompass.de/en/sources/)