Soorma.ai provides an open-source, event-driven infrastructure platform designed to support the development and management of production-grade AI agents. Positioned as a control plane within the agent stack, the platform addresses reliability and state management issues by offering framework-agnostic tools for memory and orchestration. It utilizes a NATS-powered event bus and PostgreSQL with pgvector to handle complex agent choreography, moving away from simple request-response models toward persistent, multi-threaded systems.
The company focuses on the implementation of the CoALA (Cognitive Architectures for LLM Agents) framework to resolve "AI amnesia" through episodic, semantic, and procedural memory services. For developers and SREs, Soorma is relevant because it abstracts the operational complexities of managing asynchronous messaging and vector databases. By championing a distributed cognition framework, the company aims to enable "Agentic SaaS," where agents function as stateful participants in business workflows rather than isolated chat interfaces.
Soorma is building the foundational "control plane" for the AI agent ecosystem. Their primary mission is to resolve the fragmentation and unreliability inherent in current agent development by providing an open-source, event-driven platform. They are focused on "Agentic SaaS," empowering developers to build agents that are not merely single-threaded chatbots, but sophisticated, choreographed systems with persistent state and robust memory.
The "secret sauce" lies in their transition from traditional request-response models to a Distributed Cognition (DisCo) framework. They solve the pervasive "AI Amnesia" problem through a rigorous implementation of the CoALA (Cognitive Architectures for LLM Agents) framework, providing semantic, episodic, and procedural memory. Their platform eliminates the operational friction of manually managing complex infrastructure—such as NATS, PostgreSQL, and state machines—allowing developers to focus exclusively on agent logic.
The platform functions via a "Context API" that exposes core services to agents. Developers utilize the soorma dev CLI for local iteration with hot-reloading. Agents (built using any framework) interact via a SoormaClient to publish and subscribe to business events on a NATS-powered bus. The platform automatically tracks state and persists agent memory in PostgreSQL using pgvector for semantic lookups, ensuring seamless, zero-config observability.
While specific founder identities are not yet publicly detailed, the company operates with a strong "Building in Public" ethos. The project is spearheaded by an engineering-heavy team with deep expertise in high-scale asynchronous systems, as evidenced by their NATS and PostgreSQL-centric stack. The company is currently in a pre-v1 stage, with a targeted broader launch for Q1 2026.
The primary Ideal Customer Profile (ICP) includes AI Engineers, SREs, and CTOs at enterprises and high-growth SMBs. They specifically target teams building long-running autonomous workflows in sectors like Fleet Management, Fintech, and complex B2B SaaS, where reliability, scalability, and data sovereignty (via self-hosting) are non-negotiable.
Soorma positions itself as an "Open Core" alternative to proprietary platforms like OpenAI's Assistants API or LangChain's LangGraph Cloud. It acts as a universal bridge between diverse frameworks (LangChain, CrewAI, AutoGen) and production-grade infrastructure. By focusing on event-driven choreography rather than simple orchestration, they disrupt the common, more fragile linear agent loops.
soorma dev CLI mimics the successful developer-first patterns seen in tools like Supabase or Firebase.An open-source, event-driven platform for agentic SaaS.