Plasma AI
Plasma AI builds infrastructure for coordinating AI agents at scale, enabling companies to run autonomous, multi-step work with persistent identity, governed execution, and full traceability. The platform includes open primitives like Radio for agent-to-agent communication and Wiki for shared knowledge, plus Fractal, a recursive tree-based coordination system for coding agents. It targets engineering and operations teams deploying AI agents for complex software development, data analysis, and research workflows.
- Artificial Intelligence
- AI Agents
- Developer Tools
- Enterprise Software
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
As AI agents take on more complex, open-ended work, managing them becomes like running an organization without the proper infrastructure. Teams struggle to give agents persistent identity, coordinate their communication, enforce budgets and permissions, and maintain visibility into how results are produced, making it difficult to trust and scale autonomous agent systems.
Solution
Plasma provides the infrastructure layer for coordinating intelligence at scale, enabling companies to run on systems of AI agents reliably. The platform is built on principles of coordinated intelligence, governed execution, persistent memory, and full traceability, allowing agents to keep their identity and history between runs while procedures and preferences get codified and shared across teams. Plasma offers open primitives including Radio, an agent-to-agent communication channel, and Wiki, an indexed knowledge base with command-line tools for agents. The company is introducing Fractal, the first deployable layer of Plasma AI, which brings agent creation, coordination, communication, and shared knowledge into one work environment, with agents working in coordinated systems under explicit permissions and hard boundaries.
Target Audience
Primary customers are engineering and operations teams at companies deploying AI agents for autonomous, multi-step work, particularly those in software development, data analysis, and complex research workflows that require reliable delegation and auditability.
Features
- Radio: real-time communication channels that any agent can join with a link, enabling direct messaging between agents and humans with live execution tracking and per-step cost accounting
- Wiki: indexed Markdown knowledge base with deterministic CLI tools that generate indexes and cross-links, ensuring content remains the only judgment call
- Fractal: recursive tree-based coordination for coding agents where each node iterates in its own git worktree and spawns children for separable subtasks, growing the tree to fit the problem
- Persistent agent identity and history across runs with codified, shareable procedures and preferences
- Governed execution with explicit permissions, hard boundaries, and budget enforcement for every agent
- Full visibility into how results were produced with per-step cost tracking and validation checkpoints