Starter plan

Starter is the free individual plan. It includes Perch AI Web, Desktop, and CLI under one account, a rotating list of frontier models subject to availability, and up to 20,000 PT of Perch-funded hosted usage each month. No credit card is required.

Start with Starter or compare every plan.

What Starter includes

One Perch Token (PT) represents $0.001 of metered hosted-model usage. Starter usage refreshes each month. The product shows the active allowance in Settings → Usage; in the CLI, run /usage.

Hosted usage and private inference

Starter's included allowance is Perch-funded hosted usage delivered through Perch AI Web, Desktop, and CLI. Perch's internal application endpoints are not a public API.

You can instead add your own key from an OpenAI-compatible provider or point Perch at a self-hosted endpoint. That inference is paid for and governed by you and your provider. It does not consume the Starter allowance and is excluded from the Starter Data Program described below.

Starter Data Program

The free plan includes a data program that helps support Perch-funded hosted usage. It is enabled by default after a versioned notice is shown and can be turned off at any time in Settings → Privacy.

The program applies only when all of these are true:

• the interaction is generated on Starter;
• Perch is funding the hosted model request;
• the Starter Data Program is enabled; and
• the required program notice and acknowledgement have occurred.

It does not apply to Pro, Enterprise, active or complimentary Pro access, bring-your-own-key usage, or self-hosted model usage. Eligibility is based on the plan and inference source at the time an interaction is generated. Upgrading or downgrading later does not reclassify earlier activity.

The program applies prospectively to eligible interactions generated on or after September 13, 2026. New Starter users see the program notice when selecting the plan. Existing Starter users participate prospectively only after acknowledging the notice. Historical activity is not included without separate explicit authorization.

What Derived Training Data can contain

Perch may process eligible Starter activity into Derived Training Data through filtering, redaction, abstraction, transformation, annotation, generalization, aggregation, suppression, and quality review. Derived Training Data may preserve useful task and action structure, including transformed or deidentified:

• task descriptions and model inputs and outputs;
• tool selections, tool arguments, and tool results;
• failures, retries, and recovery sequences;
• workflow steps and evaluation results; and
• approval, rejection, and correction signals.

Perch may use and commercially license Derived Training Data to develop, train, fine-tune, evaluate, benchmark, test, and improve AI models, agents, systems, datasets, and related technologies, including technologies developed by third parties.

What is always excluded

Perch does not sell or license raw User Content through the Starter Data Program. Derived Training Data excludes:

• raw prompts, chats, model outputs, tool arguments, and tool results;
• files, documents, workspace materials, and confidential source material;
• email content and browser page content;
• personal information, names, account and profile details, and payment information;
• credentials, passwords, authentication tokens, API keys, and secrets; and
• anything that remains reasonably linkable to a person, account, workspace, customer, organization, or confidential source.

A dataset may retain the structure of an email or browser tool action only after the underlying email or browser content is removed. Records that cannot be adequately transformed, or that present a material privacy, confidentiality, intellectual-property, or security risk, are excluded.

Before Derived Training Data is used outside Perch, controls are designed to remove direct and indirect identifiers, detect secrets and sensitive content, abstract source-specific details, suppress rare identifying events, deduplicate material, and assess linkability. Recipients are contractually barred from reidentification, reconstruction of User Content, combining data to identify a source, contacting affected people or organizations, and targeted advertising.

Turn the program off

Open Settings → Privacy and turn off Starter Data Program. This stops future eligible collection and use for the program. Where feasible, Perch also removes eligible records that are still raw and have not been exported or incorporated into Derived Training Data.

Completed Derived Training Data is designed so that it is no longer reasonably linkable to a source. Perch may therefore no longer be able to associate it with your account or remove it from a dataset already delivered to a recipient or from model weights already trained with it. Turning off the program does not affect processing needed to provide, secure, support, debug, meter, or enforce Perch.

Move to Pro

Pro adds every premium hosted model and larger, separate usage lanes. Activity generated while Pro is active—including complimentary Pro access—is excluded from the Starter Data Program. Earlier eligible Starter activity is not reclassified by an upgrade.

Related pages

• Pricing compares Starter, Pro, and Enterprise.
• Pro plan explains Pro allowances and data handling.
• Models lists the hosted model registry.
• Bring your own key explains private inference outside Perch-funded usage.
• Security and data explains how Perch handles files, requests, and plan boundaries.
• Privacy Policy provides the governing program details.