Gumloop

What it does

Writes code

PartlyAgents write and execute Python and shell in a per-chat cloud sandbox, and can commit files through the GitHub connector, but this is task scripting rather than repository-scale software authoring.The Code Sandbox gives your agent the ability to execute Python code and shell commands in a secure, isolated cloud environment.docs.gumloop.com/core-concepts/agent_sa… ↗

Reviews code

PartlyThrough 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)docs.gumloop.com/nodes/mcp/github ↗

Ships it

PartlyGumloop 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.docs.gumloop.com/enterprise-features/ho… ↗

Done-for-you service

NoYou buy self-serve software and build the agents yourself; the only vendor-people option is an optional Embedded Gumloop Expert on Enterprise, which the plan table lists as support, not delivery.Gumloop is an AI agent platform. Build AI agents that use your tools and data to do real work, then automate them with triggers, schedules, and the…docs.gumloop.com ↗

Not established: Debugs and fixes

How much it does unattended

Runs autonomously

YesScheduled, app and webhook triggers run agents unattended with no chat window open, and approval gates are opt-in per tool rather than mandatory.Agents can run autonomously without manual interaction.docs.gumloop.com/core-concepts/agent_tr… ↗

Multi-agent

YesAn agent can clone itself to parallelise research or hand work to named specialist agents, each running in its own conversation and sandbox.Skills teach this agent how to do a job. Subagents are other agents you hand work to.docs.gumloop.com/help/skills-triggers-m… ↗

Agent permissions

YesPer-app presets (always allow, ask each time, ask for writes and deletes, custom), per-tool never-allow, and CEL rules on the actual arguments decide what runs before it asks.The agent never executes a gated tool call without your explicit go-ahead.docs.gumloop.com/core-concepts/human_in… ↗

Which models it runs on

Claude

YesAnthropic models are selectable per agent and through the completions endpoint; the one Anthropic family with 30-day retention stays blocked until an admin acknowledges it.OpenAI-compatible chat completions endpoint, multiplexed across every model Gumloop supports (Anthropic, OpenAI, Google Gemini, OpenRouter routes).docs.gumloop.com/api-reference/chat-com… ↗

GPT

YesOpenAI is one of the providers you can browse in the agent model picker, and its models are also reachable through the OpenAI-compatible completions API.To pick a specific model instead, search by name or browse by provider: Anthropic, OpenAI, Google, DeepSeek, MiniMax, Z.ai, and more.docs.gumloop.com/core-concepts/ai_models ↗

Open models

YesOpen-weight models from DeepSeek, Z.ai, MiniMax and others run as agents, and Gumloop markets them as the cheaper default rather than a fallback.Each model is open weight and available to run as an agent in Gumloop.gumloop.com/open-source-ai-models ↗

Model choice

YesYou pin the model per agent from 35 or more options or hand routing to Auto; Enterprise admins can restrict the list and point presets elsewhere.You choose the model in Agent Preferences, at the top of the agent's configuration.docs.gumloop.com/core-concepts/ai_models ↗

Where you use it

In your editor

PartlyThere is no Gumloop editor extension; IDEs reach it as MCP clients, so Cursor or VS Code drive Gumloop agents rather than hosting them in the editor.Connect to the Gumloop MCP server from Claude, Cursor, VS Code, and other MCP-compatible clients.docs.gumloop.com/mcp-server/overview ↗

On the command line

YesA macOS and Linux CLI with a JSON mode covers agents, sessions, chat, MCP tools and artifacts; native Windows is unsupported and needs WSL or the SDK.gumloop is the command line for Gumloop. Sign in once, then drive your agents, sessions, chat completions, MCP integrations, Company Brain, skills,…docs.gumloop.com/cli/overview ↗

In your pipeline

PartlyWebhook triggers and a scriptable CLI let CI jobs or cron start an agent, but Gumloop ships no pull-request check, CI app or build-status integration of its own.Anything that can send an HTTP POST — Stripe, GitHub, a cron job, an internal script — runs the agent by calling it.docs.gumloop.com/api-reference/getting-… ↗

In a browser

YesAgents are built and chatted with in the gumloop.com web app, and any agent can also be published as a standalone hosted chat page for other people.Hosted Pages let you publish your agent on a dedicated URL at gumloopagents.com.docs.gumloop.com/core-concepts/hosted_p… ↗

Whose machine it runs on

Self-hosted

PartlyRunning Gumloop on your own infrastructure is an Enterprise VPC option that the plan table marks Optional; Pro is vendor-hosted with no self-run path.Enterprise customers running Gumloop in their own VPC are not charged for compute at all, since it runs on their infrastructure.docs.gumloop.com/core-concepts/credits ↗

Not established: Open source

What it costs to run

How it meters

YesMetering is per credit at $0.005: model tokens at provider cost, one credit per successful tool call, five credits per session-minute, plus an eight per cent orchestration fee.Credits are the currency that powers Gumloop. Every agent conversation spends credits based on the AI model it uses, the tools it calls, and how long…docs.gumloop.com/core-concepts/credits ↗

Free tier

PartlyYou can sign up without talking to sales, but the only free route is a card-backed 14-day Pro trial that converts to paid; no standing zero-dollar plan is offered.Every new account starts with a 14-day free trial of Pro. The trial requires a card, is a one-time offer per customerdocs.gumloop.com/core-concepts/credits ↗

API access

YesA REST API with key or bearer auth, unauthenticated webhook trigger URLs, and Python and JavaScript SDKs cover agent creation, sessions and completions on both plans.The Gumloop API is agent-first: you can create and update agents, start sessions to chat with them, and use chat completions from any…docs.gumloop.com/api-reference/getting-… ↗

MCP server

YesA remote MCP server lets another AI client list, create and run Gumloop agents, and the plan table checks Gumloop MCP for both Pro and Enterprise.Gumloop exposes a remote Model Context Protocol (MCP) server that lets any compatible AI client manage your agents, sessions, skills, and more.docs.gumloop.com/mcp-server/overview ↗

Bring your own key

YesYour own provider keys are accepted on both plans and zero the token charge, but the orchestration fee doubles to sixteen per cent of the notional list cost.the credits that would have been allocated toward the cost of tokens drop to zero, and the orchestration fee increases from 8% to 16%gumloop.com/pricing ↗

Buying it for a team

A company can buy it

YesAn organisation account holds billing, users and security, with unlimited seats and teams on Pro and a sales-quoted contract on Enterprise.Your company's account that manages billing, users, and securitydocs.gumloop.com/core-concepts/teams ↗

Seat model

NoGumloop is not sold per seat: the plan table lists unlimited seats and teams on both tiers, so cost tracks credit consumption rather than headcount.gumloop.com/pricing ↗

Pooled budget

YesOne organisation-wide credit pool funds every member's agents, and an API endpoint lets admins set a monthly credit limit per custom role on top of it.Everyone in your organization shares the same credit pool.docs.gumloop.com/core-concepts/teams ↗

Admin controls

PartlyAnyone can gate their own agent's tool calls, but organisation-wide app policies, role-based access and model restrictions are Enterprise-only, so Pro governance stops at the agent.App Policies is an Enterprise feature.docs.gumloop.com/enterprise-features/ap… ↗

Audit log

PartlyAudit logs record sign-ins, credential access and organisation changes, but the plan table checks the row only for Enterprise, so Pro has no administrative trail.Audit Logging provides comprehensive tracking and monitoring of user actions across your Gumloop organization.docs.gumloop.com/enterprise-features/au… ↗

Single sign-on

PartlySAML single sign-on and SCIM provisioning with Okta, Entra, Google, JumpCloud and Ping require Enterprise; Pro members sign in with Google, Microsoft or a password.SAML and SCIM settings require the Admin organization role and an Enterprise subscription.docs.gumloop.com/enterprise-features/ss… ↗

What happens to your code

Opt out of training

YesThird-party models run under zero-data-retention agreements by default, and the one family that retains prompts for thirty days stays blocked until an admin explicitly acknowledges it.Every model Gumloop serves runs under Zero Data Retention except Anthropic's Claude Fable family, where Anthropic keeps prompts and outputs for 30…docs.gumloop.com/enterprise-features/ai… ↗

Data residency

PartlyJurisdiction is only controllable on Enterprise by deploying into your own cloud account; the hosted service publishes no region choice, and its outbound traffic egresses from Oregon.Deploy Gumloop inside your own AWS, Azure, or Google Cloud environment to keep data in your network.gumloop.com ↗

Getting out

PartlyBulk export of runs, agents and audit logs is an Enterprise admin feature; everyone can pull artifacts and skills through the CLI, and Pro's monthly credits expire unused.Usage Data Export allows enterprise organization administrators to export comprehensive data from their Gumloop organization.docs.gumloop.com/enterprise-features/or… ↗

Certifications

YesThe vendor claims an independently audited SOC 2 Type II, and its trust centre lists SOC 2 Type 2, GDPR and HIPAA as compliant with reports available on request.SOC 2 Type II. Independently audited and compliant.gumloop.com ↗