platoseedplatoseed
Open-source infrastructure for reliable long-running agents
OpenProse is building open-source infrastructure for long-running AI work: a declarative language plus the Reactor harness/runtime for reliable agent workflows. Our organic open-source traction (7k+ installs) has already translated into a customer at a medical research lab, starting with a $10k pilot set to roll into a $10k/month contract. YC recruited us into the current batch and we've since hired two senior founding team members (ex-Google, Forbes 30u30 harness engineer). All three founding team members are two time founders. We just closed a $1m capped SAFE from Jonathan Abrams 8-Bit Capital and expect to raise $2.5-$3m more on Demo Day.
OpenProse is an open-source infrastructure that provides a programmable runtime for long-running AI sessions, using a Markdown-based contract system to define behavior. It orchestrates multi-agent flows by connecting a Prose language, a framework (Forme), and a Reactor runtime to execute contracts.
OpenProse uses *.prose.md sources as the authoritative contracts. A Prose program defines the AI session behavior in Markdown, including services, tools, and strategies. The Forme framework wires requires to ensures, compiling contracts into a dependency graph, and the Reactor runtime executes the runtime with content-addressed receipts and status tracking. The stack enables multi-agent coordination by mapping agents and skills through a runtime that manages orchestration, dependencies, and execution flow.
Who it’s for: Teams or organizations building AI-powered workflows that require durable, verifiable long-running AI sessions and multi-agent orchestration.
Active project with documentation, examples, and community mentions; open-source repository likely active (mentions MIT license and GitHub).
Software engineer of 12 years, agent developer of 5 years, two time founder. Creator of OpenProse.
Monitoring and learning layer for long-running agents
Enterprise AI customer support
