Frequently asked questions

AgentWorks is an AI agent platform that lets you deploy, manage, and monitor AI agents for business automation. It comes with 50+ pre-built agent templates across marketing, sales, HR, finance, support, and data categories. The platform is designed for EU AI Act and GDPR compliance from the ground up.

Sign up for free and you receive a claimable €5 welcome credit in tokens to try any agent template. No credit card is required. Pick a template from the agent directory, configure it for your workflow, and start using it immediately.

No. Agent templates are ready to use out of the box. You configure them through a visual interface — selecting models, connecting integrations, and setting approval rules. For advanced customization, API access is available on all plans.

ChatGPT and Copilot are general-purpose AI assistants. AgentWorks provides specialized, task-specific AI agents that integrate into your business workflows with compliance controls, audit logging, and multi-agent orchestration. Think of it as going from a single AI chat to a managed team of AI workers.

You pay only for the AI tokens your agents consume, billed at a published per-model rate — no opaque credit packs. Each model has a transparent per-token rate and your wallet shows real-time spend. Pro includes €10 of balance every month; Team includes €10 per seat every month (pooled at the tenant level); usage beyond it is billed from your balance, which you can top up any time. Workflow automation is metered separately as runs: every plan includes a monthly run allowance and extra runs are bought in packages. (Pro is a flat single-user plan; Team is priced per seat.)

AgentWorks offers four plans: Free (€0 — 50+ AI agents, up to 4 integrations, 100 automation runs a month, a claimable €5 one-time welcome credit after verifying your email), Pro (€39/month for individuals, or €32.50/month billed yearly, with all standard and custom agents, unlimited workflows, schedules, 2,500 automation runs a month, and priority support), Team (€49 per seat/month, or €40.83 billed yearly, for collaborating groups, adding shared chat, shared knowledge bases, team admin, and 10,000 automation runs a month), and Enterprise (custom pricing with advanced agents, unlimited runs, SSO, and an SLA). Yearly billing is two months free. Pro includes €10 of balance every month; Team includes €10 per seat every month; usage beyond it is billed from your balance at our published per-model rates (see /models).

Yes. The Free plan is €0, pay-as-you-go, and you can claim a €5 one-time welcome credit after verifying your email — enough to run multiple agent tasks and evaluate the platform. No credit card is required to sign up at app.agent-works.ai.

Token usage is billed at a published per-model rate, shown live in your balance per turn and per agent run. Workflow automation is counted in runs: each plan includes a monthly run allowance (100 on Free, 2,500 on Pro, 10,000 on Team) and extra runs come in packages from €100 for 30,000 runs. To keep spend down, the AUTO router classifies each message and routes it to the cheapest capable model — typically cutting model cost by around 30% versus always using a top-tier model. See the pricing page for per-model rates.

AgentWorks supports leading models including Anthropic Claude (Opus, Sonnet, Haiku), OpenAI GPT-5 and GPT-5 mini, Google Gemini, and Mistral. An AUTO router can pick the cheapest capable model per message, or you can select a model per agent based on task, cost, and performance.

Multi-agent pipelines let you chain multiple AI agents together in a workflow. For example, a research agent gathers information, a writing agent creates content, and a review agent checks quality — all running automatically with human approval at critical steps.

Yes. Beyond the 50+ templates, you can create custom agents with specific instructions, model selection, integration connections, and approval rules. Custom agents can be saved as templates and shared across your team.

RAG (Retrieval-Augmented Generation) connects your company documents, websites, and databases to AI agents. Instead of generating responses from general training data, agents ground their answers in your actual business information — reducing hallucinations and increasing accuracy.

AgentWorks is EU AI Act-ready rather than blanket "compliant" — under the Act, whether a use case is high-risk depends on how you deploy the agent. AgentWorks provides the building blocks the Act expects: per-agent risk classification, complete audit logging, an optional PII-detection layer, human-in-the-loop controls, and EU data residency. Every agent interaction is traceable for regulatory review.

You define approval rules per agent. When an agent reaches a critical decision point — such as sending an email, updating a CRM, or publishing content — it pauses and routes the action to a human reviewer. The reviewer can approve, reject, or modify before the action executes.

AgentWorks is fully hosted in the EU by default, on infrastructure we operate ourselves in Germany. All data is encrypted in transit and at rest. We do not use your data to train AI models.

Yes. Every agent interaction, approval decision, and data access is logged with timestamps, user identity, and full context. Audit logs are exportable and designed to meet EU AI Act traceability requirements.

AgentWorks connects to the business tools teams use day-to-day — email, Google Workspace, Slack, CRM (HubSpot and Pipedrive), Exact Online, and Notion — plus MCP (Model Context Protocol) servers for advanced connectivity with custom tools. New connectors are added based on customer demand.

No. AgentWorks is a managed SaaS with no self-host tier — every plan, including Enterprise, runs on our own infrastructure, fully hosted in the EU (Germany). Your data stays in the EU because of where and how we host it, plus GDPR-by-design controls — not because you run the platform yourself.

Yes. The AgentWorks REST API lets you programmatically create agents, trigger workflows, manage knowledge bases, and retrieve analytics. API access is available on all plans.

No. AgentWorks routes to hosted models from OpenAI, Anthropic, Google, and Mistral — there is no local or on-premise model option. Every model call runs through our managed EU-region infrastructure with zero-retention model contracts.