October 22, 2026

Grass brings three different things under one name: a network that collects public web data, rewards for people who supply internet bandwidth, and a token-delegation program.[9][17][10] Those activities connect, but they are not interchangeable. In particular, the documented Stage 2 bandwidth distribution pays USDC, while the staking program awards GRASS.[9][17][10]

That distinction is the starting point for understanding the product. Sharing a connection is not the same decision as locking tokens with a router, and useful data collection does not, by itself, establish income for every token holder.

What is the customer buying?

Grass describes its original business as collecting web data for AI training: supplying snapshots of information that model developers can use before a training run.[15] Its September 28 strategy article presents live retrieval as the next step, rather than describing every planned search product as already available.[15]

On October 8, Grass announced its Contents API, which accepts a public webpage URL and returns cleaned HTML, Markdown, or text.[2] The announcement describes JavaScript rendering, geographic targeting and freshness controls.[2] Access requests are reviewed rather than offered as unrestricted anonymous use.[2]

For a developer, the proposed deliverable is therefore retrieved page content—not an AI model, a promise of better answers, or ownership of the source material.[2] Grass describes escalating from lightweight retrieval to more capable infrastructure and, where needed, a browser environment.[2]

Consider an application that needs a current product description. Getting readable content from the page is one job; deciding whether that description is accurate is another. The API addresses the retrieval step. Its launch is evidence of a defined product offer, not proof that a particular customer’s application works well.

Grass also publishes retrieval benchmarks, but those are its own tests.[2][15] This review did not independently repeat them, inspect paying-customer contracts, or reconcile commercial revenue. The company’s claims about serving leading AI laboratories should be read as vendor claims, not independently established adoption.[2][15]

Why residential bandwidth matters

The network’s node documentation describes a device relaying web requests and returning encrypted web-server responses to a router.[9] Traffic allocation depends on factors including reputation and geographic demand.[9]

In plain language, a participant supplies a route to the public internet.[15] Grass’s strategy is to use residential connections to reach information that conventional AI crawlers struggle to retrieve.[15]

This makes the location and usefulness of a connection commercially relevant, not just the fact that an application is switched on.[4] The points documentation distinguishes availability from actual use and says stability, geographic demand and previous routing activity can affect whether a connection receives requests.[4]

That is a more useful mental model than treating every spare connection as equally valuable. A large participant count would describe potential supply; it would not tell us how much customers bought, whether they renewed, or what margin remained after operating costs. Those are separate questions.

Points are not a token balance

Grass documents two contribution measures.[4] Uptime Points reflect connection availability and also include certain referral and milestone rewards.[4] Network Points reflect bandwidth actually used to fulfil web requests.[4] The documentation says referrals do not generate Network Points.[4]

Neither label should be read as a standing exchange rate into GRASS. The relevant distribution announcement must specify what participants can receive and which past activity qualifies.

The July 10 Stage 2 announcement covers activity from October 14, 2024 through June 8, 2026.[17] It specifies USDC rewards, more than zero Network Points as an eligibility condition, and January 22, 2027 as the claim deadline.[17] It also says holding GRASS has no effect on that distribution and that this distribution creates no new GRASS emissions.[17]

These are rules for a named historical contribution period, not an earnings forecast for someone joining on October 22. Nor should its published conversion rates be carried forward into an assumed future payout schedule.

The rewards terms separately give Grass control over distribution eligibility and identify the official dashboard as the claim venue.[7] Eligibility restrictions still apply.[7]

The important evidence boundary is simple: a documented claim program is stronger than an unexplained points counter, but it is not an independently reconciled payout record. This review did not connect a wallet, submit a claim, or verify completed participant transfers.

What GRASS staking actually does

Staking is a separate use of the token.[10] The program’s terms define it as delegating GRASS to an available router operated by Grass or an approved third party.[10] They define staking rewards as GRASS awarded to routers and, through them, eligible delegators.[10]

A router operator retains its commission; the balance is divided among its delegators in proportion to their stake with that router.[10] That is a share of the router’s awarded staking rewards—not a documented percentage of the entire company’s sales.[10]

For a beginner, the sequence is best understood as four decisions, not a promise of passive income:

  • Choose the delegation relationship. Grass’s explainer says its available-router list can change and that third-party approval assesses technical capabilities, not the operator’s finances or conduct.[1]
  • Understand the estimate. Displayed APR is variable, not a commitment to future rewards.[1]
  • Account for administration. Rewards require manual claiming and redelegation to compound; transactions incur gas fees.[1]
  • Understand the exit. An unstaking request cannot be cancelled.[10] The terms specify seven days of locked access and forfeit rewards on those tokens that have not yet been awarded when the request is submitted.[10] Grass reserves the ability to change that period.[10]

The seven-day wait should not be confused with a requirement to remain staked for a minimum earning term.[8] The short staking documentation explicitly distinguishes no minimum staking period from the unstaking delay.[8]

There is also a documentation trap.[8][10] That short page describes rewards as distributed every second; the terms warn that rewards may be delayed or never received.[8][10] A description of distribution cadence is not a guarantee of payment to an individual delegator.[8][10]

Grass’s explainer says slashing is not currently part of the program but could be introduced, and that a router can be barred from participation, affecting rewards.[1] Those statements describe published rules, not a fresh audit of deployed contracts.[1]

Does customer demand create token demand?

The economic question is not simply whether AI companies need data. It is how a customer payment travels through Grass and what, if anything, a token holder is entitled to receive.

Grass’s October 2 structure article says subsidiaries handle different operations: OpCo runs the protocol and associated programs, DataCo supplies public web data, and SearchCo develops live-retrieval products.[3] It says those businesses receive customer revenue and are wholly owned by the Grass Foundation.[3]

That is useful organizational disclosure. It does not, by itself, grant GRASS holders a claim on subsidiary profits. Keeping revenue within a foundation-controlled structure and distributing that revenue to token holders are different propositions.

The tokenomics documentation also identifies an initial router-incentive pool intended to support infrastructure until network fees can sustain it.[13] Accordingly, staking rewards should not automatically be labelled customer-funded yield: an incentive allocation and fees from customers are different funding sources.[13]

Likewise, the Contents API announcement explains the service but does not establish a mandatory GRASS purchase at customer checkout or reconcile conversions into tokens.[2]

The cautious conclusion is not that GRASS has no role. Delegation and token rewards are explicit. What remains unestablished here is the complete, currently executed connection between incremental customer revenue and benefits for token holders. This review has not reconciled fee-funded rewards, token purchases, circulating supply changes or distributions against financial records.

A persuasive economic account would need to separate customer receipts, operating expenses, incentive-token allocations and actual payments to participants. More network traffic alone cannot fill those gaps.

Privacy and safety: separate the promises

Grass’s node documentation says the network routes traffic through a participant’s IP address without accessing their private activity, and describes the collected material as public web data.[9] These are claims about the intended separation between network requests and personal browsing.[9]

They do not mean Grass collects no personal information.[6] Its privacy policy lists account information, wallet addresses and IP addresses, and also describes device information, analytics and disclosures to service providers.[6]

That distinction matters: not reading a user’s private browsing is narrower than collecting no identifying information. This review did not inspect application binaries, extension permissions or live traffic, so it cannot certify those privacy claims through hands-on testing.

Grass’s network-use policy describes business verification, approved use cases and prohibited activities such as hacking, phishing and unlawful circumvention.[11] It also describes monitoring and enforcement measures.[11]

Those controls are relevant, but publishing a policy is not proof that every request complies with it.[11] Similarly, a page being publicly reachable does not settle every question about how its content may be collected or reused.[11] Grass itself requires compliance with applicable law and contractual restrictions.[11]

For a prospective participant, the practical questions remain: Are you authorized to share this connection? What do your network provider’s terms permit? Are bandwidth allowances or other costs acceptable? What permissions does the actual software request? A reward counter is not an answer to those questions.

The useful dividing line

Grass is best evaluated as a data-access business with several participation mechanisms, not as a single earnings product.[2][4][10] Its documentation supplies concrete explanations of web retrieval, contribution measurement and router delegation.[2][4][10]

Keep three tests separate: whether customers find the delivered data useful, whether bandwidth participation makes sense under the applicable terms, and whether token delegation’s commission, uncertainty and exit restrictions are acceptable. Evidence for one does not settle the others.

For someone considering staking, understanding what is awarded, who takes a commission and when tokens become accessible matters more than a headline APR. For someone considering bandwidth sharing, the crucial distinction is between points credited now and a specifically defined reward distribution later. Neither decision needs a bullish token-price thesis.

This is a product and mechanism explainer, not investment advice. It is based on public documentation checked on October 22, 2026, without installing Grass, sharing bandwidth or transacting.

Sources