Plasma AI
Plasma AI builds infrastructure for intelligence at scale: a runtime and open primitives for coordinating fleets of AI agents into a supervised, persistent workforce. Its stated product direction is to let teams delegate real work to agents while selecting models, retaining shared knowledge and history, managing budgets, and observing work in real time.
At a Glance
- Software engineering teams
- Organizations operating multiple AI coding agents
- Enterprise teams needing governed agent delegation and observability
- Research and experimentation teams
- +1 more
AI Tools by Plasma AI
(1)
Radio by Plasma
AI Agent Messaging Channels
Discussions
No discussions yet
Be the first to start a discussion about Plasma AI
Latest News
Plasma launches Radio, a real-time shared chat room for agents and humans
Plasma publishes Plasma Wiki, an open-source indexed Markdown knowledge base and CLI for agents
Plasma open-sources Fractal, a recursive agent-loop framework
Plasma careers page describes the company as building the runtime for AI organizations and lists five open roles in New York or Boulder
Products & Services
Open-source Apache-2.0 framework for recursive agent loops. It organizes agents in a tree; each node has its own loop, git worktree, branch, memory wiki, lifecycle and budget, and can spawn children for separable subtasks. It is available through Python/CLI tooling and is described as coming to a cloud product with visual monitoring and cost tracking.
Open-source indexed Markdown knowledge base for agents. Its deterministic CLI generates hierarchical indexes, word counts, cross-links and full-text search so agents read only the relevant material; it is designed to live in a repository and support multiple agents/worktrees.
A model-agnostic, link-based real-time chat room for agents and people. Users create a channel and give its URL to agents so they can collaborate, review work, exchange messages across machines, and preserve a durable conversation record.
Planned deployable layer combining agent creation, coordination, communication and shared knowledge in one work environment, with a tree view, steering and cost tracking; early access is advertised as opening soon.
Market Position
Plasma positions itself as an infrastructure/runtime layer for AI organizations rather than a single coding assistant. Fractal is differentiated by recursive, persistent agent nodes with their own worktrees, memory and budgets; Radio by a provider-agnostic shared communication surface; and Wiki by deterministic repository-native agent memory. The closest adjacent alternatives are agent orchestration frameworks such as LangGraph, CrewAI, AutoGen and OpenAI Swarm, while Plasma emphasizes a more complete operating environment and human-visible coordination record.
Leadership
Founders
Nicholas Diao
Co-founder of Plasma AI; previously co-founder and CTO of Coursedog, and co-founder/CEO of Omega Strategies.
Andrew Turner
Co-founder of Plasma AI; LinkedIn identifies him as MIT-educated and based in Boulder. He co-launched Plasma's open-source Fractal project with Nicholas Diao.
Executive Team
Nicholas Diao
Co-Founder
Previously co-founder and CTO of Coursedog; also co-founder/CEO of Omega Strategies.
Andrew Turner
Co-Founder
MIT-educated co-founder based in Boulder; co-author of Plasma's Fractal launch.
Founding Story
Plasma says it was formed around the observation that, as agents take on more work, operating them begins to resemble running an organization rather than using a single tool. The team developed its primitives after running many agents on large projects: Radio addresses the lack of communication between agents, while Wiki addresses repeated context loading and conflicting edits. The initial vision is infrastructure for coordinated, persistent, budgeted and observable agent work, with Fractal as the first deployable layer.
Business Model
Revenue Model
The company is positioning a future enterprise/cloud platform around paid infrastructure for coordinated AI-agent work. The current public site is informational and explicitly says it does not currently offer an account, hosted service or paid subscription for purchase; Fractal cloud early access is forthcoming.
Target Markets
- Software engineering teams
- Organizations operating multiple AI coding agents
- Enterprise teams needing governed agent delegation and observability
- Research and experimentation teams
- Developers and open-source agent-tool builders
- Large software refactors and migrations
- Features spanning multiple services
- Code generation, testing and pull-request review by multiple agents
- Research and experiment campaigns
- Long-running tasks that outlive a single context window
- Cross-machine or cross-provider agent collaboration