Problem

OpenProse addresses the reliability gap in long-running AI-agent work. Conventional agent workflows are often imperative scripts: users specify step one, step two, and step three, then manually wire dependencies and hope the run produces the intended result. As workflows become more complex, changes can require reworking the sequence and rerunning unchanged work. OpenProse’s economic proposition is to reduce wasted model execution and operator attention by making workflows repeatable, incremental, and auditable.

The clearest use case is running recurring business operations as standing AI responsibilities. OpenProse’s own marketing, go-to-market, and operations reportedly run on responsibilities for mention tracking, competitor intelligence, lead enrichment, outreach drafting, and market indices. These are ongoing jobs where the world changes incrementally and reliability matters more than a one-off agent demo.

Product / Service

OpenProse is an open-source declarative language and runtime for AI work. Users write familiar Markdown contracts in .prose.md files, declaring what a responsibility requires and what truth it maintains rather than prescribing every execution step. The system compiles those contracts into a dependency graph, uses Forme to wire compatible inputs and outputs, and runs them on a compatible agent harness such as Claude Code, Codex, OpenCode, or Pi. Reactor is the project’s recommended deterministic harness.

At runtime, Reactor maintains a world model, watches for changes, and re-renders only the facets whose inputs have moved, while memoizing agent sessions in a dependency graph. This is intended to make long-running workflows more reliable and portable, while durable traces, world-model state, and signed receipts provide an audit trail for decisions and execution.

Market

OpenProse competes in open-source AI-agent infrastructure, specifically the emerging category of programming languages and runtimes for reliable, long-running agent workflows. Its adjacent alternatives include LangChain, CrewAI, and AutoGen, which orchestrate agents externally, as well as BAML and DSPy, which provide other programming or orchestration layers. OpenProse differentiates itself by putting the workflow’s declarative contracts and dependency resolution inside the AI-session model rather than treating orchestration as an external configuration layer.

The company appears to be at an early commercialization stage. Y Combinator lists OpenProse as an active Spring 2026 company founded in 2026 with a three-person team in San Francisco, and a company profile reports $500,000 raised from Y Combinator. Its strongest disclosed traction is product usage: the company says its own marketing, GTM, and operations run on thirteen OpenProse responsibilities. No public revenue figure or customer revenue is established in the retrieved evidence, so it is more accurate to describe OpenProse as early-stage with revenue undisclosed than to assert definitively that it is pre-revenue.