Overview

The Grass network is designed to operate as a decentralized infrastructure that collects, verifies, and structures publicly available web data for use in AI development. At its core, the architecture includes three main roles: Grass Nodes, Routers, and Validators. Each plays a specific function in the flow of bandwidth, data, and verification. Grass Nodes are operated by users who voluntarily share their unused internet bandwidth. Routers coordinate requests and responses between users and data endpoints, while Validators are responsible for verifying the integrity of these interactions and committing them to the blockchain through cryptographic proofs.

This layered system ensures both scalability and verifiability. Rather than forcing each node to independently verify and broadcast data on-chain—which would be inefficient—Grass uses Validators to verify batches of interactions using zero-knowledge proofs. These proofs confirm that a particular action (such as a web request) was completed correctly without revealing the actual content of the data or the identity of the user. This method helps maintain privacy while still providing on-chain accountability, an essential balance in any bandwidth-sharing or data-mining protocol.

A key benefit of Grass’s architecture is the separation of roles. Users can contribute bandwidth without needing to run a full node or participate in complex verification processes. Routers specialize in managing communication pathways and optimizing traffic flow. Validators focus on verifying correctness, constructing zero-knowledge proofs, and ensuring only valid data is rewarded. This separation prevents bottlenecks and allows each part of the network to scale according to its specific performance demands.

The system is also designed to be modular and upgradeable. While some initial components—like Validators—are run by the Grass Foundation or trusted parties, the long-term plan includes opening up these roles to the community through staking, governance, and open-source development. In time, anyone will be able to run a Router or Validator, subject to performance criteria and bonding requirements, creating a more trustless and decentralized system.

Validator

Validators in the Grass network are tasked with maintaining the integrity and reliability of bandwidth usage by verifying traffic relayed through Routers and submitted by Grass Nodes. When data is sent or received through the network, Validators ensure that it meets protocol standards, was delivered as expected, and complies with any quality constraints. To do this, they use zero-knowledge proof systems that confirm data activity without needing to expose user information. These proofs are then recorded on-chain, serving as immutable evidence of work done and bandwidth shared.

Initially, the Validator layer is operated in a semi-centralized manner by the Grass Foundation. This is a deliberate design choice during the early stages of the network, as it allows for stable testing, security monitoring, and performance calibration. However, the roadmap includes a transition to a decentralized validator committee model. In this future phase, Validators will be elected or chosen through a staking mechanism, allowing anyone who meets the hardware and protocol requirements to participate.

Validators are required to manage large volumes of data and generate complex cryptographic proofs efficiently. To do this, they rely on specialized infrastructure capable of high-throughput computation. This layer is less about bandwidth and more about computation and security. Each Validator must maintain a history of routing logs, track Router performance, and detect any anomalies or signs of abuse. For example, if a Router is repeatedly sending incomplete requests or serving invalid responses, the Validator flags it and can reduce its reward eligibility or reputation.

Validators also serve as the gatekeepers for protocol rewards. Only bandwidth activity that has been verified and confirmed by a Validator is eligible for token issuance. This step ensures that only honest, high-quality participants receive compensation, thereby reducing fraud and misuse. The integrity of the staking and airdrop systems relies heavily on the Validator’s ability to process data accurately and impartially.

Router

Routers act as the coordination layer between users (Grass Nodes) and the broader Grass network. Their primary role is to receive tasks from the Validator layer and distribute them to connected Grass Nodes based on performance, reliability, and location. In essence, Routers act as intermediaries that facilitate the flow of data requests and ensure that bandwidth contributions are properly routed to fulfill specific scraping or retrieval goals. They don’t collect or verify the data themselves but are responsible for tracking the performance of the Nodes they connect and ensuring traffic is relayed accurately and efficiently.

Each Router maintains a set of metrics about the Nodes under its management. These include latency, packet loss, availability, and task completion rate. This performance data is used to determine how much traffic each Node should receive and plays a role in reward calculations. Routers also generate reports that are passed on to Validators, who assess whether the traffic was completed correctly and whether token rewards should be issued. This constant performance monitoring makes Routers an essential link in the Grass incentive system.

To incentivize reliable Router operation, Grass allows token delegation to Routers. This mechanism functions similarly to proof-of-stake models in other networks: Routers with more delegated GRASS tokens receive more traffic, increasing their earning potential. This encourages router operators to maintain high uptime and stable performance. Delegators, in turn, earn a share of the router’s rewards, creating a shared interest in the router’s success. This structure creates competition among Routers, which improves network reliability and decentralization over time.

Routers can set their own commission rates, which affects how much of their rewards they keep versus what is passed on to delegators. Users looking to stake GRASS tokens must evaluate routers based on past performance, uptime statistics, and commission rates to make informed delegation decisions. This adds a reputational layer to Router participation and ensures the network continuously rewards high-quality operators while deprioritizing underperformers.

Grass Node

Grass Nodes are the primary access point for individual users to participate in the Grass protocol. By installing the Grass browser extension or desktop application, users allow their devices to share unused internet bandwidth, which the network uses to make web requests for publicly available content. These requests may involve retrieving webpage data, API responses, or media files needed for AI training. The Grass Node software runs in the background and is designed to operate passively without requiring technical involvement from the user.

From a security perspective, Grass Nodes are configured to only handle traffic that is considered safe and publicly accessible. The protocol does not interact with private networks, password-protected websites, or personal user data. All requests are filtered to ensure they comply with safety and legal guidelines. The Grass team has implemented encryption and basic sandboxing measures to isolate the Node’s activity from other processes on the host device. This is supported by independent certifications from cybersecurity vendors and antivirus companies to verify that the software poses no threat to the user’s system.

Nodes receive tasks from Routers based on availability, bandwidth capacity, and reputation. The more consistently a Node performs, the more requests it will be assigned. Nodes that experience downtime, fail to complete requests, or deliver incomplete data will gradually receive fewer tasks and thus fewer rewards. This reputation-based distribution system ensures that the network prioritizes reliable participants while still allowing anyone to join and improve their standing over time.

User rewards are calculated based on the volume and quality of bandwidth contributed. Points are awarded for each verified request completed, which can later be used to claim GRASS tokens through airdrops or direct reward systems. This structure allows users to earn passively from their existing internet connection without needing to buy tokens or interact with exchanges. In future updates, the Grass Node may also support staking or additional participation modes that give users more control over how their bandwidth is used.

Traffic Types

The Grass protocol is built around collecting and structuring public web data that can be used to train AI systems. To achieve this, the network must process a wide variety of traffic types that reflect the diversity of content found on the internet. Traffic on the Grass network includes requests for HTML pages, static images, structured data (such as JSON from APIs), media files, and metadata. By enabling this broad range of traffic, the network creates a robust and flexible dataset that can serve different machine learning use cases—from language models to image classification and recommendation systems.

Each traffic type is categorized and handled based on its characteristics. For example, text-based data (like HTML or JSON) is parsed and stored in structured formats suitable for language model training. Image and media traffic, on the other hand, is processed differently to ensure efficient transfer and categorization based on resolution, file type, or origin domain. This classification is important because it allows AI developers to request only the data types relevant to their training pipeline, reducing noise and improving processing efficiency downstream.

Grass Nodes are responsible for executing the web requests that retrieve this data. Routers assign tasks to Nodes based on location, performance, and compatibility with the required traffic type. For instance, a Node with faster internet speeds and lower latency might be assigned heavier tasks such as image or video retrieval, while smaller Nodes might focus on lightweight API calls. This dynamic matching system allows the network to distribute traffic efficiently and ensure each Node contributes according to its capacity.

Not all traffic types carry the same value. Some are more computationally expensive, require more bandwidth, or have higher demand from data buyers. To reflect this, the Grass network assigns different point values to each type of traffic. More demanding or high-priority tasks earn higher rewards, while simpler ones earn less. This system helps balance incentives and ensures the network remains efficient under varying workloads. It also encourages users with better hardware or faster connections to take on higher-value roles within the network.

Fee Market

Grass introduces a fee market to regulate demand and prioritize traffic in a decentralized environment. Unlike fixed-rate systems, the Grass fee model dynamically adjusts based on the characteristics of each task and current network conditions. The formula for calculating transaction fees includes five main variables: geography (g), reputation (r), traffic type (t), bandwidth used (b), and network congestion (c). Each variable contributes to a weighted total that determines the final fee to be paid for the request, ensuring that costs reflect the complexity and resource demands of the operation.

The geography variable (g) reflects the location of the Grass Node handling the request. Some tasks may require region-specific data, such as content that is only accessible within a given country. In these cases, Nodes in the relevant geographic area are prioritized and the associated fees are adjusted upward to reflect the scarcity and value of that access. This mechanism helps direct traffic intelligently while rewarding participants in regions that are underrepresented or in higher demand.

The reputation variable (r) reflects the performance history of the Node. Nodes with a strong track record of uptime, low latency, and task completion will receive better-paying requests and lower risk of rejection. Conversely, Nodes with weaker performance will have higher fee thresholds and may be deprioritized in routing decisions. This approach links long-term behavior directly to earning potential, giving users a clear reason to maintain stable and compliant participation over time.

The traffic type (t) and bandwidth (b) variables are tied to the technical nature of each request. As covered in the previous section, different traffic types have different demands in terms of data volume, complexity, and value. A lightweight API call will be cheaper than downloading a full-resolution image gallery, and the fee model reflects this accordingly. Bandwidth consumption is also tracked precisely, with larger requests costing more both for the requester and in terms of token rewards issued to the contributing Node.

Network congestion (c) is the final variable in the formula and helps the system self-regulate during periods of high demand. When congestion rises, fees adjust upward to prioritize only the most urgent or valuable traffic. This prevents overload, maintains request reliability, and ensures that routers and validators are not overwhelmed by low-priority tasks. The congestion multiplier may be adjusted in future versions to reflect real-time usage patterns and optimize performance.

Grass Reputation Scoring

To ensure data quality and network reliability, Grass implements a reputation scoring system that measures Node performance over time. Each Node is evaluated based on four key characteristics: Completeness, Consistency, Timeliness, and Availability. These metrics are recorded automatically as Nodes handle traffic and are combined into a weighted reputation score. This score directly influences how much traffic a Node receives, what types of tasks it’s eligible for, and how it is prioritized in the fee market.

Completeness refers to whether a Node successfully delivered the expected content for a request. If a webpage is only partially loaded or an API response is truncated, the request is marked incomplete. This affects the Node’s score and can reduce future earning opportunities. Grass uses automated validation tools, often supported by the Validator layer, to confirm whether a response meets the completeness criteria before approving rewards or counting the request toward reputation.

Consistency measures how reliably a Node delivers accurate data across repeated requests. A high-performing Node will consistently return correct and expected responses, even when tasks are repeated or randomized for auditing purposes. This metric is especially important in filtering out unreliable Nodes or those attempting to manipulate the system with spoofed results. Consistency checks are performed periodically and are factored into long-term reputation assessments.

Timeliness evaluates the latency and speed of each request. A Node that consistently responds quickly to data tasks is seen as more reliable and receives a higher score. Nodes with slower response times or frequent timeouts are penalized in reputation. Since the Grass network is used to collect data in near real-time for AI model training, responsiveness is crucial. Timeliness scoring helps ensure that users running Nodes are maintaining stable connections and that the network can be used in high-throughput applications.

Availability tracks the uptime of a Node—how often it is online and ready to receive traffic. Nodes that frequently disconnect or remain inactive for extended periods lose standing in the reputation system. Conversely, Nodes that are reliably online over long periods are rewarded with higher-tier traffic and improved earning potential. Availability is especially important for Router operators and heavy contributors who wish to run dedicated hardware or provide continuous uptime.

Highlights

• Grass relies on a three-layer architecture where Nodes share bandwidth, Routers handle request distribution, and Validators generate zero-knowledge proofs to confirm data validity.
• Validators verify traffic correctness and log proofs on-chain, ensuring that only completed and accurate data requests result in rewards.
• Routers receive delegated stake, manage Node performance, and direct traffic based on geography, speed, and historical reliability.
• Grass Nodes execute public web requests while protecting user privacy, and their participation is rewarded based on volume, quality, and consistency of bandwidth shared.
• The protocol uses a reputation system and fee market, which together prioritize high-performing Nodes and dynamically adjust reward structures based on traffic type, location, and congestion.