> For the complete documentation index, see [llms.txt](https://policies.appfarm.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://policies.appfarm.io/service-and-usage/agent-action-node-pilot-terms.md).
# Agent Action Node Pilot Terms
These terms apply to the Customer's use of the Agent Action Node in Appfarm Create during its pilot phase (the "Pilot"). The Agent Action Node is a new action node that lets the Customer configure AI agents in its Solutions: the Customer selects a model, describes a task, and grants the agent a set of tools, and the agent then reasons about the task and calls those tools — reading and writing data, running actions and flows — to complete it.
The Agent Action Node is not yet generally released. Appfarm is making it available to selected customers ahead of launch so they can start building with it early, at Appfarm's cost price for model consumption, in exchange for accepting that the functionality is in beta and that the Customer is responsible for its use. These terms set out that arrangement. They supplement the Customer's Subscription Agreement as a specific agreement for new functionality under section 12.1 of the Software Service Agreement, and apply to all use of the Agent Action Node in any of the Customer's Solutions and Environments during the Pilot Period.
**Acceptance.** The Agent Action Node is enabled per Solution on request. Before it is enabled, the Customer accepts these terms in writing (email is sufficient) through a representative authorized to bind the Customer. Appfarm records who accepted and when.
## Pilot status
The Agent Action Node is functionality under development in the meaning of section 12.2 of the Software Service Agreement. In practice, this means:
* Functionality, configuration options, available models, and limits may change, be added, or be removed during the Pilot Period, with or without notice. Agents built during the Pilot may need to be reconfigured, and configurations may not carry over unchanged to the released version.
* The warranty and the Service Level Agreement do not apply to the Agent Action Node, or to errors, unavailability, or data changes in a Solution caused by it. Appfarm nonetheless welcomes bug reports and feedback at `olav@appfarm.io` and will address them at its discretion.
* Appfarm may suspend, limit (including rate-limit), or discontinue the Agent Action Node — for all customers, for a specific model, or for a specific Solution — at any time where Appfarm considers this necessary for operational, security, cost, or compliance reasons, or where a model provider requires it.
## Pilot Period
The Pilot Period runs from the Customer's acceptance until Appfarm notifies the Customer that the Pilot has ended for the Customer, either because the Agent Action Node moves to general release on commercial terms (section 3) or because Appfarm withdraws it. The Customer may stop using the Agent Action Node at any time by ceasing to use it or asking Appfarm to disable it, and then owes only the consumption charges incurred up to that point. The Customer accepts that they will make themselves available to provide feedback on their use of the Agent Action Node.
## Charges during the Pilot, and what follows
Appfarm has not finalized the commercial model for the Agent Action Node.
During the Pilot Period, the Customer is charged only Appfarm's cost for the model consumption of its agents, without any additional markup from Appfarm. This is invoiced in arrears in accordance with the [Invoicing and Payment Policy](https://policies.appfarm.io/service-and-usage/invoicing-and-payment-policy). Agent Action Node consumption is not covered by, and does not consume, Appfarm AI Credits.
When the commercial model is ready, Appfarm will notify the Customer in writing. Use of the Agent Action Node after that date is subject to the commercial terms; the Customer may stop using the Agent Action Node before that date with no further obligation. Commercial charges are never applied retroactively to consumption during the Pilot Period.
## The Customer controls the agent — and is responsible for what it does
An agent can only do what the Customer configures it to do. The Customer chooses the model, writes the system and task prompts, selects which data sources, actions, and flows the agent may call, and which operations it may perform on them, and sets the safety limits (iterations, timeout, failure thresholds). Appfarm does not automatically control, review, or verify the agent's reasoning, the tool calls it makes, or their results.
Appfarm may assist the Customer during the Pilot, for example, with designing, configuring, and testing agents. Such assistance is part of the Pilot and does not shift responsibility to Appfarm: the Customer reviews and approves each agent configuration before it is used, and remains responsible for it and for the agent's actions as set out below.
The Customer therefore acknowledges and accepts that:
* Agent behavior is probabilistic. The same configuration may produce different results across runs, and the results may be incomplete, inaccurate, or not what was intended. Outputs and actions must not be relied upon without verification appropriate to the use case.
* Actions performed by an agent, including creating, updating, and persisting data, deleting data, sending messages, and calling external systems, take effect immediately and are not rolled back when a run completes, fails, or hits a safety limit. Restoring data changed by an agent is not a warranty incident (Appfarm may assist under its Backup Policy).
* The Customer is responsible for all actions performed by agents in its Solutions, including their effects on the Customer's data, users, and third parties, and for ensuring that those actions comply with applicable law and the Customer's own obligations.
Before using the Agent Action Node for anything consequential, the Customer should test configurations in development and test Environments before deploying to production, grant an agent only the tools and operations its task requires, keep a human review step in front of irreversible actions, set conservative safety limits, and monitor runs using the outputs and logs the node provides. These are recommendations, not conditions, but the risk of not following them rests with the Customer.
## Data and privacy
When an agent runs, the Customer's runtime data (prompts, the content of the data sources the agent can access, and the results of its tool calls) is sent to the provider of the model selected for that agent. This differs from Appfarm AI in Appfarm Create, which only processes development prompts and application metadata. Appfarm provides information about available models and their providers as part of the Pilot (made available in the Agent Action Node through a model selection function), and the Customer selects or approves the models that its agents use. By enabling the Agent Action Node and using a model, the Customer instructs Appfarm to transmit runtime data to that model's provider to run the agent.
The Customer remains the controller of any personal data processed by its agents under the Data Processing Agreement, and is responsible for having a lawful basis for that processing and for deciding which data, in particular data classified as Personal or Sensitive in the Customer's data model, it exposes to an agent. Use of the Agent Action Node is subject to Appfarm's Acceptable Use Policy and to the usage policies of the selected model provider.
## Liability
The Agent Action Node is provided as is during the Pilot, at cost and without warranty. Appfarm accepts no liability for loss or damage arising from the Agent Action Node or from actions performed, or not performed, by agents in the Customer's Solutions, including loss of or changes to data, incorrect or incomplete outputs, unavailability, and third-party claims. The Customer holds Appfarm harmless from third-party claims arising from the Customer's agents. Section 12 of the Software Service Agreement otherwise applies. For the avoidance of doubt, this section applies to the Agent Action Node only; the Customer's rights under the Subscription Agreement for the rest of the Service are unaffected.
## Confidentiality and feedback
Information about the Agent Action Node that Appfarm has not made public, including roadmap, pricing indications, and known limitations, is Appfarm's confidential information under the Subscription Agreement. The Customer may use its agents freely, but shall not publicly announce its participation in the Pilot without Appfarm's consent, which will not be unreasonably withheld. Feedback, suggestions, and bug reports may be used by Appfarm to improve the Service without restriction or compensation.
## Relationship to the Subscription Agreement
These terms supplement the Customer's Subscription Agreement and apply to the Agent Action Node only. All other terms of the Subscription Agreement and Appfarm's policies remain unchanged and in full effect. Appfarm may update these terms during the Pilot Period with notice to the Customer in accordance with section 13.3 of the Software Service Agreement.
## Revision history
| Date | Revision |
|---|
16.09.2026 | Document published. |
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://policies.appfarm.io/service-and-usage/agent-action-node-pilot-terms.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.