Unlimited users. Unlimited agents and workflows. Pay only when your agents run.

An agent run is any tracked execution that creates a trace. Sending a message to an agent, creating an annotation, or starting an agent from a schedule or event each creates one agent run and one trace. Each new message, annotation, or trigger is a separate run.

An agent run is the customer-facing usage unit. A trace is the technical record Agenta creates for that run so you can inspect its model calls, tool calls, timing, outputs, and errors. Every agent run creates one trace.

Yes, on every paid Cloud plan and both self-hosted plans. Invite as many team members as you need and create as many projects, agents, workflows, schedules, and triggers as you want. The free Cloud Hobby plan includes two team members. You pay based on agent runs, not seats, projects, agents, steps, or workflows.

No. Connect your own model-provider credentials and pay the provider directly.

Connect OpenAI, Anthropic, Gemini, Mistral, Groq, MiniMax, Together AI, OpenRouter, Azure OpenAI, AWS Bedrock, or Google Vertex AI. You can also connect an OpenAI-compatible endpoint.

Hobby stops running agents after 5,000 agent runs in a monthly billing period. Upgrade to Pro to continue. Pro and Business include 10,000 agent runs per month. Additional runs are billed automatically at $5 per 10,000, and your agents continue running without interruption.

No. Included agent runs reset at the start of each monthly billing period.

Open Settings, then Billing to see your current agent-run usage and plan allowance.

Count each message sent to an agent, each annotation, and each invocation started by a schedule or event. For example, one daily scheduled agent uses about 30 or 31 runs per month, one agent triggered every hour uses about 720 to 744 runs per month, and a chat with 20 user messages uses 20 runs.

Trace data retention controls how long Agenta Cloud keeps technical execution records, including model calls, tool calls, timing, outputs, errors, evaluations, and annotations. Hobby retains trace data for one week, Pro for one month, and Business for three months. Enterprise retention is configurable. Trace data retention does not describe the retention of chats, project files, or persistent agent context. With self-hosted Agenta, you manage trace storage and retention in your own infrastructure.

Agenta Cloud is operated by us. We run the application, storage, agent runner, upgrades, and backups. With self-hosted Agenta, you deploy and operate those components in your own infrastructure, connect your own model and integration accounts, and control where your data is stored.

The open-source edition includes the core agent workspace: chat, projects, agent configuration, instructions, skills, tools, integrations, MCP servers, files, schedules, event triggers, tracing, evaluations, version history, role-based access control, and SSO. You can create unlimited users, projects, agents, and workflows on your own infrastructure.

Enterprise adds audit logs, custom domains, deployment and upgrade support, priority support, a private Slack Connect channel, dedicated support, and a custom service-level agreement. The open-source edition remains free for teams that can operate it themselves and do not need those commercial features or services.

Yes. Upgrade, downgrade, or cancel from Settings, then Billing.

Yes. Early-stage startups, students, and researchers can email team@agenta.ai to ask about discounted or free access. Include what you are building and how you plan to use Agenta.