# Gumloop — Multiplayer AI agent builder: teams build agents over their apps while IT keeps governance and spend control.

> Source: The Agents Index — https://theagentsindex.com/gumloop (structured, researched, re-verified)
> Facts last verified: 2026-09-30

Gumloop is a hosted platform where non-technical and technical teams build AI agents that read and write their company's apps through 100+ MCP connectors, triggers and skills. Agents run on 35+ models and produce artifacts such as dashboards, documents and live web apps.

| Fact | Value |
| --- | --- |
| Website | https://www.gumloop.com |
| Price | from $37/month |
| Free trial | Partly |

14-day free trial of Pro, started self-serve from the pricing page; no standing $0 plan appears in the published plan table (only Pro and Enterprise). Card requirement not stated on the pricing page.

## What it does

**Writes code** · Yes — Agents generate and execute Python and shell code in a per-conversation cloud sandbox that is on by default, and can ship code as a shareable artifact.

> Give an agent a prompt and Gumloop agents can write, execute, and iterate on their code in a sandboxed environment.
> — https://www.gumloop.com/vs/n8n · 'What Gumloop does that n8n can't' section, 'Agents that write and run their own code'

**Reviews code** · Partly — Through the GitHub connector an agent can open a pull request's changed files and inline review comments and post its own comment, but there is no dedicated review product or verdict workflow.

> List inline review comments on a PR (code-level comments on specific lines/files)
> — https://docs.gumloop.com/nodes/mcp/github · GitHub connector page, Available Tools table, List Pull Request Review Comments row

**Ships it** · Partly — Gumloop builds and deploys custom MCP servers straight from a GitHub repository on every push to main, but that ships the connector code, not the buyer's own applications.

> You write the code in a GitHub repository, and Gumloop handles building, deploying, monitoring, and routing requests to your server.
> — https://docs.gumloop.com/enterprise-features/hosted_mcps · Hosted MCPs, opening paragraph

**Done-for-you service** · No — Gumloop sells self-serve software your own team builds on; the only people-led option is an optional embedded expert on Enterprise, not a done-for-you build service.

> Embedded Gumloop Expert
> — https://www.gumloop.com/pricing · Plan comparison table, Support section, 'Embedded Gumloop Expert' row (Pro: no, Enterprise: Optional)

_Not established: Debugs and fixes, Writes tests_

## How much it does unattended

**Runs autonomously** · Yes — Plain-English triggers let agents monitor apps or the web and act with no human in the loop; scheduled and webhook triggers run unattended.

> Tell Gumloop in plain English what to watch for on the web or across any combination of apps, and it'll monitor and take action automatically.
> — https://www.gumloop.com/products/agents · 'Agents that run themselves' feature block

**Multi-agent** · Yes — Multiple agents can be orchestrated inside deterministic workflows and an agent can delegate to subagents, so coordination is a first-class feature.

> Orchestrate complex processes with multiple agents inside deterministic workflows. Agents can even call on subagents.
> — https://www.gumloop.com/products/agents · 'Coordinate teams of agents' feature block

**Agent permissions** · Yes — Per-connector approval modes range from always-allow to never-allow, with a read-free write-gated middle setting and per-tool custom rules.

> "Ask for writes/deletes" is a great default for most apps. It lets your agent gather information freely while keeping you in the loop for any changes.
> — https://docs.gumloop.com/core-concepts/human_in_the_loop · Setting Up Approval Controls, App-Level Settings tab, tip below the approval mode table

## Which models it runs on

**Claude** · Yes — Anthropic is a first-class provider in the model picker and the docs list nine Claude models served directly by Anthropic.

> To pick a specific model instead, search by name or browse by provider: Anthropic, OpenAI, Google, DeepSeek, MiniMax, Z.ai, and more.
> — https://docs.gumloop.com/core-concepts/ai_models · Choosing a model section, provider list (Available models table lists Anthropic rows)

**GPT** · Yes — OpenAI is a listed provider and the available-models table carries several GPT entries served through OpenAI directly.

> To pick a specific model instead, search by name or browse by provider: Anthropic, OpenAI, Google, DeepSeek, MiniMax, Z.ai, and more.
> — https://docs.gumloop.com/core-concepts/ai_models · Choosing a model section, provider list (Available models table lists OpenAI rows)

**Open models** · Yes — Open-weight families are runnable as hosted agents, but always through Gumloop's US-based inference providers rather than weights you host yourself.

> All open-source models (such as GPT-OSS 120B, LLaMA, DeepSeek, Qwen, Kimi, MiniMax, and GLM) are accessed through US-based providers under Zero Data Retention (ZDR) policies
> — https://docs.gumloop.com/core-concepts/ai_models · Available models section, note beneath the provider table

**Model choice** · Yes — The buyer picks the model per agent from 35+ options on both plans, can swap freely, and Enterprise admins can restrict the allowed set.

> You choose which models and providers your agents use, and you can bring your own keys (BYOK) so calls run through your own provider accounts.
> — https://www.gumloop.com/solutions/security · 'Privacy and how your data is used' section

## Where you use it

**In your editor** · Partly — There is no Gumloop IDE extension; instead its remote MCP server is documented for Cursor, VS Code and Windsurf, so the editor drives Gumloop agents rather than Gumloop living in the editor.

> Connect to the Gumloop MCP server from Claude, Cursor, VS Code, and other MCP-compatible clients.
> — https://docs.gumloop.com/mcp-server/overview · MCP Server page, introduction and Setup subsections

**On the command line** · Yes — A first-party CLI runs agents, calls MCP tools, searches Company Brain and manages skills and artifacts from the terminal, and is included on both plans.

> Sign in once, then run agents, call MCP tools, search Company Brain, and manage skills and artifacts from your terminal.
> — https://www.gumloop.com/products/cli · Page hero beneath the 'Gumloop CLI' heading

**In your pipeline** · Partly — Webhook triggers let GitHub, a cron job or an internal script start an agent, but the vendor documents no CI-native runner or agent that acts on a pull request.

> Anything that can send an HTTP POST — Stripe, GitHub, a cron job, an internal script — runs the agent by calling it.
> — https://docs.gumloop.com/api-reference/getting-started · Agent Webhooks section, first paragraph

**In a browser** · Yes — Gumloop is a hosted web product and agents additionally get standalone public pages, plus Slack, Teams and email surfaces.

> Give your agent a standalone page on gumloopagents.com that anyone can open directly.
> — https://docs.gumloop.com/core-concepts/hosted_pages · Hosted Pages entry description in the documentation index

## Whose machine it runs on

**Self-hosted** · Partly — Virtual Private Cloud deployment is an optional Enterprise line item only, and Gumloop operates the stack inside your cloud rather than handing you software to run.

> Run Gumloop as managed SaaS, or deploy into your own cloud (VPC), in the region of your choice.
> — https://www.gumloop.com/solutions/security · 'Deployment you control' section

_Not established: Open source_

## What it costs to run

**How it meters** · Yes — Billing is credit-metered on top of a flat plan fee: model tokens, compute and paid tool calls at list price, plus an 8% orchestration fee, with credits priced at half a cent.

> Credits are indexed at $0.005 each. Pro includes 7,400 + 12,600 bonus credits (20,000 total) for $37/month.
> — https://www.gumloop.com/pricing · Frequently asked questions, 'What is a credit?'

**Free tier** · Partly — There is a self-serve 14-day free trial of Pro but no standing free plan in the published table, so continued use requires the $37/month tier.

> Start 14-day free trial
> — https://www.gumloop.com/pricing · Plan comparison table header, Pro column call-to-action

**API access** · Yes — A documented REST API with Python and JavaScript SDKs creates agents, starts sessions and triggers automations, and is listed for both plans.

> The Gumloop API is agent-first: you can create and update agents, start chat sessions, and use chat completions from any OpenAI-compatible client.
> — https://docs.gumloop.com/api-reference/getting-started · Getting Started, first paragraph

**MCP server** · Yes — Gumloop runs a remote MCP server exposing dozens of tools so another agent can list agents, run sessions and search Company Brain; MCP is a Pro-and-above line item.

> Gumloop exposes a remote Model Context Protocol (MCP) server that lets any compatible AI client manage your agents, sessions, skills, and more.
> — https://docs.gumloop.com/mcp-server/overview · MCP Server page, opening paragraph

**Bring your own key** · Yes — Your own provider keys make model calls cost zero credits, but the orchestration fee doubles to 16% and each key only waives the models that provider serves.

> BYOK makes AI model calls cost 0 credits, and it also applies to image generation and voice transcription. Since the model spend leaves your credit balance entirely, agent chats on BYOK carry a 16% orchestration fee instead of 8%
> — https://docs.gumloop.com/core-concepts/ai_models · Bring your own key (BYOK) section

## Buying it for a team

**A company can buy it** · Yes — An Enterprise plan is sold through sales with custom pricing, org-wide governance and SCIM provisioning, alongside self-serve team workspaces on Pro.

> We work with our Enterprise teams on custom plans so that your team can scale confidently with an extremely obvious ROI.
> — https://www.gumloop.com/pricing · Frequently asked questions, 'How much does Enterprise cost?'

**Seat model** · Yes — Seats and teams are unlimited on both plans with no per-seat charge or stated minimum, so cost scales with credit consumption instead of headcount.

> Unlimited
> — https://www.gumloop.com/pricing · Plan comparison table, Collaboration section, 'Seats' row, Pro and Enterprise columns

**Pooled budget** · Yes — Team members draw on one shared credit balance on both plans rather than each holding a private allowance, which is what makes unlimited seats workable.

> On both Pro and Enterprise plans, team members can share credentials, collaborate on agents, and use the same pool of credits.
> — https://www.gumloop.com/pricing · Frequently asked questions, 'Can I collaborate with others?'

**Admin controls** · Partly — Pro limits connector guardrails to a single agent's scope; org-wide policies, role-based access and model access control are Enterprise-only rows in the same table.

> Agent-scoped
> — https://www.gumloop.com/pricing · Plan comparison table, Admin & Security section, 'Connector Policies & Guardrails' row, Pro column (Enterprise column reads 'Org-wide')

**Audit log** · Partly — Audit logs are marked absent on Pro and present on Enterprise, so administrative traceability requires the sales-led plan.

> Our Enterprise tier offers the observability, security, and compliance features that large teams require, like RBAC, organization-wide governance policies, audit logs, incognito mode, and more.
> — https://www.gumloop.com/pricing · Frequently asked questions, 'Why upgrade to Enterprise?'; Admin & Security table 'Audit Logs' row

**Single sign-on** · Partly — SAML single sign-on and SCIM provisioning work with any identity provider but the plan table marks SCIM/SAML support Enterprise-only.

> Connect any identity provider for single sign-on (SSO) and automated user provisioning with SCIM.
> — https://www.gumloop.com/solutions/security · 'Access and infrastructure' section (pricing table 'SCIM/SAML Support' row: Enterprise only)

## What happens to your code

**Opt out of training** · Yes — Zero-data-retention agreements with major model providers are the default rather than a setting, and open models are explicitly excluded from training; a 30-day-retention model group must be chosen deliberately.

> Your data is never used for model training and is not stored after inference.
> — https://docs.gumloop.com/core-concepts/ai_models · Available models section, note on open-source models beneath the provider table

**Data residency** · Partly — Region choice comes with the optional Enterprise VPC deployment; on managed SaaS the vendor names no region options and routes open models through US-based providers.

> Run Gumloop as managed SaaS, or deploy into your own cloud (VPC), in the region of your choice. Your data stays in the infrastructure you control.
> — https://www.gumloop.com/solutions/security · 'Deployment you control' section

**Getting out** · Partly — Organisation-wide data export and continuous drains are an Enterprise feature, and unused credits never carry over to the next month.

> No. The 20k credit allotment is a monthly grant that resets each billing period. You get the full 20k again next month; you don't bank what you didn't use.
> — https://www.gumloop.com/pricing · Frequently asked questions, 'Do unused credits roll over to the next month?'; Admin & Security table 'Data Exports' row

**Certifications** · Yes — SOC 2 Type II attestation, HIPAA with BAAs on eligible plans, a GDPR-aligned programme and EU-U.S. Data Privacy Framework certification including the UK extension.

> We are SOC 2 Type II attested and HIPAA compliant, and we offer BAAs for eligible plans.
> — https://www.gumloop.com/solutions/security · 'Certifications and assessments' section

## Provenance

Every fact above comes from the vendor. Nothing here is independently corroborated yet.

## Coverage elsewhere

None recorded. Every fact above comes from the vendor and is not independently corroborated yet.

## What changed

- 2026-09-30 — Capabilities · Commercial terms · Free tier
- 2026-09-26 — Free tier
- 2026-09-09 — Capabilities · Commercial terms · Free tier · Price

## Correcting this record

Anything marked "Not established" is a question we have not answered — it is not a No. If you represent Gumloop, or anything here is wrong or out of date, you can supply it with a source: the operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalogue needs no key to read, and each entry states its own auth.

Supplying evidence is free. Nothing that can be bought here changes what a field says, whether a gap is closed, or how this record is scored.