Is CrewAI Free? The Open-Source Framework vs the AMP Basic Plan

Yes, and in two separate ways. The open-source CrewAI framework is released under the MIT license, so you can install it and run crews and flows on your own machine or servers without paying CrewAI. CrewAI AMP (Agent Management Platform), the hosted product, has a free Basic plan with 2 automations and 50 workflow executions a month. Neither route makes your agents free to run. Your model provider still bills every LLM call, and paid tool APIs and your own hosting stay on you. Free software, metered inputs.

One name, two products, two kinds of free

The confusion behind the question comes from one brand covering two things. The framework is the Python package crewai: agents, tasks, crews and Flows that you define in code or JSON and run yourself. AMP is the platform on top of it, which builds crews in Studio, deploys them, triggers them on schedules or app events, monitors them and governs who can touch them. The framework is open source under MIT; AMP is CrewAI's hosted platform, with a free Basic plan. Which free version matters depends on whether you want to operate agents yourself or have CrewAI host them.

The MIT license is about as permissive as software licenses get. It lets you use, copy, modify, merge, publish, distribute, sublicense and sell the code free of charge, as long as the copyright notice stays in place ^[1]. That covers commercial work: you can build a client product on CrewAI and charge for it. Running locally needs no CrewAI account at all; crewai install locks the dependencies and crewai run starts the crew from the project root. Our CrewAI tutorial walks through the first project from install to first run.

What the free Basic plan on AMP includes

Basic is a plan, not a trial. CrewAI lists it as Free behind a Sign Up button, while the Request trial button belongs to Enterprise. Basic gives you 2 automations and 50 workflow executions a month on CrewAI cloud, and the maximum is also 50: additional executions are not included on Basic ^[2]. By our arithmetic, that is about 1.7 executions a day on a 30-day month. The ceiling is fixed. More volume lives on Enterprise, where executions are sized to the workflow with flexible overage and the price is quoted on request.

The free tier is generous on features and strict on volume. Studio, the private agent and tools repositories, workflow templates, GitHub integration, cron scheduling, the usage dashboard, token counts and workflow chat in Slack or Teams all sit on Basic. What Basic withholds is governance and scale: SSO, role-based access control, enterprise connectors, a dedicated VPC and anything beyond community support. PII redaction for traces also needs the Enterprise plan. Enterprise fees follow an Order Form, and our CrewAI pricing breakdown covers its renewal notice and the non-refundable fees rule.

What open source does not pay for

Free software is not a free bill. Every agent step calls a language model through your own provider key, and that provider bills the tokens. The same holds on AMP: CrewAI's deployment guide says you bring your own LLM API keys and any tool keys, such as Serper, so model costs are billed by those providers ^[3]. Many catalog tools also need paid third-party keys, including Tavily, Exa, Firecrawl, Bright Data and Browserbase. Serper has a free tier, and You.com search allows 100 free queries a day. A self-run crew also needs servers, and those are yours to provide.

The defaults point at OpenAI in more places than the agents. Agents use an OpenAI model unless you pick another provider. Memory embeddings default to OpenAI, knowledge embeddings default to OpenAI even when your agents run on another provider, and crew planning calls the default OpenAI model unless you set planning_llm. The official first-flow tutorial needs an OpenAI key whatever model the agents use. Switching providers therefore means changing several settings, not one. Check each before the first real run.

A fully local route exists. Ollama can run the model, connected either through the optional LiteLLM extra or as an OpenAI-compatible endpoint with custom_openai and a base_url, and an Ollama embedder keeps memory local with no external API. That removes the token bill and moves the cost into hardware you run and maintain. For tool choice, our CrewAI tools guide separates the tools that work without a key, such as the website scrapers and file readers, from the ones that need a paid account.

Telemetry: free does not mean silent

The open-source package sends anonymous usage telemetry by default ^[4]. It covers versions, crew and agent metadata, tool usage counts, LLM attributes, task lifecycle and CLI create, run and deploy attempts, and it records exception class names, never error messages. Tool names and agent roles are collected, so keep personal information out of them. Turn it off by setting CREWAI_DISABLE_TELEMETRY, OTEL_SDK_DISABLED or CREWAI_DISABLE_TRACKING to true, 1, yes or on. Any other value leaves it on. Your own OpenTelemetry spans stay yours, because CrewAI telemetry runs on its own tracer provider.

The share_crew setting goes much further, and it is opt-in. Turning it on sends complete crew information to CrewAI, including goals, backstories, task descriptions and outputs, which may contain personal information, to improve the library and train models. Leave it off for client work. Tracing is a separate switch, controlled with tracing=True, CREWAI_TRACING_ENABLED or crewai traces enable|disable. On the first run a prompt asks whether to share the trace; declining, timing out or having no prompt discards the local buffer and nothing is uploaded. Disabling tracing does not stop telemetry, and OTEL_SDK_DISABLED does not stop tracing.

When the free framework needs an AMP account

Running crews and flows on your own machine needs no account. An AMP account comes in for specific jobs. Built-in tracing needs only a free one (sign up, then crewai login). An account is also needed for deploying with crewai deploy, publishing to the Tool Repository and publishing skills to the CrewAI registry. AMP app integrations through apps=[...] need an AMP account with an active subscription, as each integration page says, plus the CREWAI_ENTERPRISE_ACTION_AUTH_TOKEN variable. Since August 2026 a project_id can also link open-source usage to an enterprise account. Each of these moves data or workloads onto CrewAI's side. That brings the platform's terms into play.

Those terms deserve one read before anything sensitive goes up. CrewAI's Terms let it use Customer Data, including AI outputs, to improve its products and services, including training models, and no opt-out is printed. Separately, the Privacy Policy says CrewAI never uses Google user data for advertising or for training AI models. Support follows the plan: Basic gets community help only, and one reviewer notes that open-source support runs mainly through asynchronous community channels rather than live assistance. For a free tier, that is a fair trade. Just know it going in.

Free to own, metered to run

CrewAI is genuinely free where it counts for builders: the MIT framework costs nothing to use, modify or sell, and Basic lets you deploy 2 automations without an Order Form or payment. What is never free is the work the agents do. Plan for a model provider bill from the first run, check which tools need paid keys, and switch off telemetry on day one if your crews touch client data. Move to Enterprise when you need more than 2 automations or 50 workflow executions a month, enterprise connectors, SSO, role-based access, a dedicated VPC or dedicated support.