AI solutions delivered through MCP infrastructure

What is SOMA?

SOMA is a marketplace for MCP (Model Context Protocol) services, built for AI agents that need to integrate, coordinate, and execute reliably at scale.

In neurobiology, the soma integrates incoming signals before triggering action. Our MCP servers play the same role for AI systems, aggregating inputs, managing state, and coordinating execution across a distributed network.

The first service live on SOMA is Context Compression, a layer that reduces token consumption while preserving output quality. Lower costs, faster responses, and more usable space inside every context window. More MCP services follow.

Challenge

Rising AI costs, shrinking returns

Token bills scale with every prompt. Context windows fill up fast. AI agents hit a wall where performance suffers and operating costs grow

Solution

MCP marketplace

SOMA delivers MCP services that cut token consumption, reduce costs, and speed up response times, starting with Context Compression.

MCPMCPMCPAI AgentSOMA

Why SOMA?

Most AI infrastructure relies on centralized teams and fixed pipelines. SOMA works differently: a decentralized, competitive network that continuously optimizes itself.

The outcome for teams building with AI agents: better performance, lower costs, and simple integrations that ship in days, not quarters.

Documentation

SOMA runs as a decentralized competitive network. Independent contributors build and submit their own implementations, and only the highest-performing solutions earn rewards.

Performance is continuously tested and ranked in an open environment. With each competition cycle, new winners can emerge and replace existing solutions. Quality evolves on its own, without disrupting the stable production layer that clients rely on.

While model improvement happens inside the competitive layer, production deployment stays controlled and predictable, delivering enterprise-grade reliability at the integration edge.

SOMA scales horizontally with demand. Capacity expands as workloads grow, keeping performance consistent for businesses of any size.

SOMARIZER

Context compression layer built on the SOMA subnet. Cut token costs, preserve meaning. Be among the first to test it.

No spam. Early access only.

Built by Dendrite

SOMA is developed and operated by Dendrite - a technology company founded in 2022, that entered the Bittensor ecosystem early and has since grown into its primary infrastructure architect.

Led by a team of 50+ elite engineers and mathematicians, Dendrite operates at every layer of the ecosystem - from designing high-performance mining infrastructure to launching proprietary subnets and building end-user products like SimplyTao.

For builders and subnet owners

Or email us directly at:

thesoma@dendrite.holdings

FAQ

SOMA is a marketplace for MCP (Model Context Protocol) services, built for AI agents that need to integrate, coordinate, and execute reliably at scale. The first service live on SOMA is Somarizer, with more MCP services in the pipeline.

MCP stands for Model Context Protocol, an open standard for connecting AI models with external tools, data sources, and execution environments. It gives AI agents a consistent way to exchange context and take action across different systems. SOMA delivers MCP services as a continuously optimized layer, so teams plug in performance instead of building it from scratch.

Context Compression is the process of reducing the number of tokens an AI model needs to produce a given output, while preserving the quality of that output. Fewer tokens translate directly into lower costs, faster responses, and more usable space inside any context window. For production AI workloads, the savings compound with every request.

Right now, SOMA has one live tool: SOMARIZER - a context compression service for AI agents and LLM applications. It plugs into existing pipelines and shrinks input tokens without compromising downstream task performance. More tools are in the works and coming soon.

SOMA | MCP Marketplace built on Subnet 114