AutoGPT
Significant Gravitas · Agent Platform With Chat-Built Agents, a Visual Builder, and Self-Hosting
AutoGPT is an agent platform where you describe a goal in chat (AutoPilot) or connect blocks in a visual builder, and the resulting agents run on schedules, webhooks, or on demand across 200+ integration blocks. It is for individuals and teams automating recurring work who want either a hosted cloud plan or a free self-hosted install with their own API keys.
Best for
Individuals and small teams who want to automate recurring research, content, and operations work with agents built through chat or a visual builder, with the option to self-host
Not ideal for
Teams that need multi-user workspaces and admin controls today, buyers who need flat and predictable costs, or companies that want to resell or host the platform as a service
Who it's for
Individuals, operators, and small teams automating recurring business workflows with no-code or chat-built agents, in the cloud or self-hosted
Capabilities
- AutoPilot chat builds an agent from a plain-English description, asks clarifying questions, and runs, debugs, and edits agents
- Dry-run self-repair loop that simulates newly generated agents and fixes errors before they run
- Visual drag-and-drop builder with sub-agents as blocks, live inputs and outputs per node, and agent import and export
- 200+ integration blocks, including Google Workspace, Slack, Discord, GitHub, Linear, Notion, HubSpot, browser automation, and MCP tools
- Cron schedules, generic webhooks, and GitHub event triggers, with cost shown per task and per block
- Human-in-the-loop approval gates that pause an agent for review and then resume it
- Experts: persistent agents with their own skills and integrations that handle repeating or one-off standing work
- Marketplace of community agents and skills, plus import of existing automations from n8n, Make, and Zapier
Limitations
- The platform code is under PolyForm Shield 1.0.0, a source-available license rather than an open-source one: it allows any use except providing a product that competes with AutoGPT (AutoGPT Classic and code outside the platform folder are MIT)
- Cloud automation runs are billed pay-as-you-go from a credit wallet on top of the subscription, and block costs vary with tokens, provider fees, and run time, so spend is hard to predict
- The pricing page lists no free cloud plan (a no-signup demo is offered), and the Team plan with multi-user workspaces is still marked coming soon
- Platform releases are still tagged beta (autogpt-platform-beta-v0.8.1)
- Self-hosting currently means a from-source Docker Compose setup (the one-command Linux and macOS installer is not live yet; on Windows the docs recommend Docker with WSL 2), you provide API keys for every model provider and service, and support is community-only
- Running self-hosted AutoPilot on a local model disables the extended-thinking path and needs a model with tool-calling support; the docs note each turn can take minutes on CPU-only hosts
Use cases
- Describing a weekly competitor-research report in chat and letting AutoPilot build, test, and schedule the agent
- Triggering an agent from a GitHub webhook to triage new issues or summarize pull requests
- Moving an existing Zapier or n8n automation into AutoGPT to add LLM reasoning steps and approval gates
- Running a lead-enrichment flow that pulls contacts from Apollo, researches them on the web, and updates HubSpot
- Setting up an Expert with its own integrations to handle recurring social posting or inbox summaries
- Self-hosting the platform with Docker and a local model so agent data stays on your own infrastructure
AutoGPT pricing
Pro
$50/month or $42.50/month billed annually
Billed monthly
- Standard AutoPilot chat allowance
- Visual builder and chat-based agent creation
- File-aware agents
- Scheduled and event-based triggers
- Email support
Max
$320/month or $272/month billed annually
Billed monthly
- 8.5x Pro's AutoPilot chat allowance
- 5x file storage
- Early access to features
- Priority support and onboarding
Team
Contact sales (coming soon)
- Multi-user workspaces
- Admin controls and roles
- Centralized billing and seat management
Self-hosted
Free
- Deploy with Docker
- Bring your own API keys for models and services
- Admin tools for the operator
Note: On the cloud, the subscription sets your AutoPilot chat allowance, while automation runs on every plan are pay-as-you-go from a prepaid credit wallet with optional auto-refill. Block costs vary by block and usage (tokens, provider cost, run time), and the platform shows an estimate before a run. Yearly billing saves 15%. Self-hosting has no license fee.
Technical specs
Available models
AutoGPT vs. competitors
AutoGPT vs. Dify
Dify is a visual platform for building LLM apps, workflows, and RAG pipelines to publish as apps or APIs; AutoGPT centers on a chat assistant that builds and runs agents for you and on persistent Experts doing scheduled standing work.
AutoGPT vs. Lindy
Lindy is a hosted, Slack-native AI teammate you delegate inbox, meeting, and follow-up work to in plain language; AutoGPT is an agent builder with a visual canvas, a marketplace, and a self-hosted option.
AutoGPT vs. CrewAI
CrewAI is a code-first framework (with a hosted platform) for multi-agent crews; AutoGPT targets people who would rather build agents through chat or drag-and-drop than write Python.
Frequently asked questions
What is AutoGPT today?
AutoGPT is now mainly the AutoGPT Platform: a hosted and self-hostable service where you create agents by describing them to AutoPilot or by connecting blocks in a visual builder, then run them on schedules, webhooks, or on demand. The original standalone agent is still available as AutoGPT Classic.
Is AutoGPT open source?
Not under an open-source license. The code is public on GitHub and AutoGPT calls the self-hosted platform open source, but the platform (the autogpt_platform folder) uses PolyForm Shield 1.0.0, a source-available license that permits any use except providing a product that competes with AutoGPT. The README summarizes it as free for personal and internal business use. AutoGPT Classic and other code outside that folder are MIT-licensed.
How much does AutoGPT cost?
The cloud Pro plan is $50/month, or $42.50/month billed annually, and Max is $320/month, or $272/month billed annually, with 8.5x Pro's AutoPilot chat allowance. On both plans, automation runs are paid separately from a pay-as-you-go credit wallet. The Team plan is coming soon, and self-hosting is free with your own API keys.
Can I self-host AutoGPT?
Yes, for free. Self-hosting currently uses a from-source Docker Compose setup (on Windows, the docs recommend Docker Desktop with WSL 2), and a one-command installer for Linux and macOS is announced but not live yet. You provide your own API keys for model providers and services, support is community-only, and AutoPilot can be pointed at a local model through any OpenAI-compatible endpoint such as Ollama or vLLM.
Does AutoGPT have an API?
Yes. The external API lets you execute agents, list blocks, and work with the store and integrations, authenticated with API keys or OAuth 2.0 with scoped permissions.
Integrations & fit
Alternatives to consider
LangGenius
No-code agentic workflows and RAG pipelines
Freemium
Lindy
Individual professionals and small teams who want an AI assistant that works across their actual email, calendar, and CRM without building explicit automation flows
PaidCrewAI
Orchestrating autonomous agent teams for enterprise tasks
FreemiumAbout AutoGPT
AutoGPT started as the well-known open-source autonomous agent, which lives on as AutoGPT Classic under the MIT license. Today the main product is the AutoGPT Platform. The main entry point is AutoPilot, a chat assistant that asks clarifying questions, builds an agent from your description, runs a simulated dry run to catch and fix errors, and then runs, debugs, or edits the agent. Behind it is a drag-and-drop builder where agents are graphs of blocks: LLM calls from all major model families, 200+ integrations (Google Workspace, Slack, GitHub, HubSpot, Notion, and more), browser automation, MCP tools, and sub-agents, with human-in-the-loop approval gates anywhere in the flow. Agents run on cron schedules, generic or GitHub webhooks, or manually, and every task shows its cost per block. Existing automations can be imported from n8n, Make, and Zapier, and a marketplace offers community-built agents. Releases in September 2026 added Experts, persistent agents with their own skills and integrations that take on repeating or one-off standing work, along with a skill marketplace and voice mode for AutoPilot. The cloud uses a two-part billing model: the Pro and Max subscriptions set your AutoPilot chat allowance, while automation runs on every plan are pay-as-you-go from a credit wallet, with block costs that vary by tokens, provider fees, and run time. Self-hosting via Docker is free, but you supply every API key, and AutoPilot can even run on a local model through an OpenAI-compatible endpoint. Two caveats: the platform code is under PolyForm Shield, a source-available license that bars using it for a competing product, rather than an open-source license, and its release tags are still labeled beta.
Updates from AutoGPT
AutoGPT Platform v0.8.1 added eight generalist Experts with skills and routines, nested workspace folders, and multi-select answers in AutoPilot's question cards.
v0.8.0 introduced hireable Experts with standing work routines (repeating or one-off), shareable Expert pages, a marketplace for publishing and installing knowledge skills, AutoPilot voice mode, and support for newer models from OpenAI, Anthropic, Google, and others. Self-hosters must follow a migration because encryption secrets are now generated per install.
Are you the founder? Claim this listing →