# Macrocosmos > Decentralising intelligence, building on Bittensor Macrocosmos builds decentralized AI infrastructure on Bittensor through competitive subnets that incentivize open-source intelligence development. Macrocosmos helps AI developers build decentralized intelligence through subnet networks on Bittensor. Users can access language models, collect data, perform protein folding, and fine-tune models. Macrocosmos operates five specialized subnets with the aim to democratize AI by making machine learning resources accessible through blockchain incentives. - Type: Company - Sector: Developer Tooling - Status: Active - Founded: 2024-01-01 - Profile: https://thegrid.id/profiles/macrocosmos ## Products - [Constellation](https://thegrid.id/profiles/macrocosmos): Constellation implements unified API access through SDK integration, enabling developers to interact with decentralized AI subnets. The platform provides client libraries for accessing language models, data collection services, and scientific computing resources across the Bittensor network. (Developer Tooling, Live, Main Product) - [Apex](https://thegrid.id/profiles/macrocosmos): Apex is a conversational AI through subnet 1 infrastructure, enabling agentic workflows with tool access and chain-of-thought reasoning. The system provides OpenAI-compatible endpoints while maintaining decentralized model execution through miner competition. (Developer Tooling, Live) - [Macrocosmos SDK](https://thegrid.id/profiles/macrocosmos): Macrocosmos SDK implements developer tools through Python and TypeScript libraries, enabling integration with decentralized AI subnets. The toolkit provides client interfaces for Apex, Gravity, and Mainframe services with standardized API access patterns. (Developer Tooling, Live) - [Mainframe](https://thegrid.id/profiles/macrocosmos): Mainframe implements protein folding computation through subnet 25 infrastructure, enabling molecular simulation and scientific computing tasks. The system coordinates distributed computing resources for biological research and drug discovery applications. (Compute Network, Live) - [Fine-Tuning](https://thegrid.id/profiles/macrocosmos): Fine-Tuning implements model optimization through subnet 37 infrastructure, enabling specialized model adaptation and performance enhancement. The system provides competitive mechanisms for improving model capabilities through targeted training approaches. (Developer Tooling, Live) - [Pre-Training](https://thegrid.id/profiles/macrocosmos): Pre-Training implements large language model training through subnet 9 infrastructure, enabling distributed model development and optimization. The system coordinates computational resources for training state-of-the-art language models through competitive incentive mechanisms. (Developer Tooling, Live) - [Gravity](https://thegrid.id/profiles/macrocosmos): Gravity enables real-time data collection and dataset delivery through Subnet 13, providing businesses with fresh, relevant data for analysis and decision-making. (OnChain Data API, Live) ## Links - [Main](https://www.macrocosmos.ai) - [Documentation](https://docs.macrocosmos.ai) - [Blog](https://macrocosmosai.substack.com) - [Whitepaper](https://www.macrocosmos.ai/research/pretraining_whitepaper.pdf) - [Telegram](https://t.me/macrocosmosai) - [Discord](https://discord.gg/macrocosmosai) - [GitHub](https://github.com/macrocosm-os) - [LinkedIn](https://linkedin.com/company/macrocosmosai) - [Twitter / X](https://x.com/MacrocosmosAI) ## Legal This data is provided under two licensing models. Model 1 - open data. The open data core is licensed under the Open Database License (ODbL) — a copyleft/share-alike license allowing free use, modification, and sharing for any purpose. Attribution is required: include "Powered by The Grid" linking to https://thegrid.id wherever the data is used. Derivative databases must be shared under ODbL terms. Trademarks, logos, and brand names (marked "Special TPIP") are not covered by the ODbL. For ML/AI use: training datasets that substantially extract ODbL data are Derivative Databases and must be shared under ODbL if publicly used; models must be attributed in documentation. Model 2 - commercial licenses. The Grid Data Service (TGDS) is also available as a commercial license. See [Web Services Terms](https://about.thegrid.id/legal/web-services-terms) for commercial use that is not covered by the open data service. The following applies to both licensing models. - All data is provided AS IS with no guarantees of accuracy. - The Grid does not provide investment, legal, or tax advice, and - a project's presence in The Grid's data does not imply endorsement. - [Data Licensing & Disclaimers](https://about.thegrid.id/legal/data-licensing-and-disclaimers): Licensing models and disclaimers - [Open Data (ODbL)](https://about.thegrid.id/legal/open-data): Open Database License terms and scope for The Grid's open data service. - [Attribution Guidelines](https://about.thegrid.id/legal/open-data-attribution-guidelines): How to attribute The Grid's data when using the open data service. - [Web Services Terms](https://about.thegrid.id/legal/web-services-terms): commercial Terms for using data services. - [All Legal Documents](https://about.thegrid.id/legal): Full index of terms, policies, and disclosures ## Optional - [TGS Field Descriptions](https://github.com/The-Grid-Data/workspace-public/blob/main/references/tgs-field-descriptions.md): Schema field definitions for The Grid's data model - [TGS JSON Structure](https://github.com/The-Grid-Data/workspace-public/blob/main/references/tgs-json-structure.md): Data model structure and relationships - [Product Type Definitions](https://github.com/The-Grid-Data/workspace-public/blob/main/references/product-type-definitions.md): Definitions for each product type - [Product Type Classification](https://github.com/The-Grid-Data/workspace-public/blob/main/references/product-type-classification.md): Guide for classifying products into types - [Grid MCP Guide](https://github.com/The-Grid-Data/workspace-public/blob/main/grid-search/skills/grid-query/references/grid-mcp-guide.md): How to query The Grid programmatically via MCP