AgentRun is a JSON/TypeScript DSL from Parcha-ai for turning repeatable AI-agent work into orchestrated, typed workflows.
Yes, AgentRun is open source under the Apache-2.0 license.
AgentRun is free to use.
Yes, AgentRun can be self-hosted (the source is available under the Apache-2.0 license).
AgentRun is a workflow DSL from Parcha-ai — expressed in JSON or TypeScript — for turning repeatable AI-agent work into orchestrated, typed workflows. It ships beta npm packages and is self-hostable under Apache-2.0. It is brand-new: created 2026-09-23, only 20 stars, and explicitly beta, so treat it as an early bet rather than a settled orchestration layer.
Most agent work in a GTM team starts as a one-off prompt and never becomes a repeatable process — the enrichment run, the outbound-draft pass, the lead-scoring loop each get rebuilt by hand. AgentRun's pitch is the orchestration layer for that: express the repeatable agent process once as a typed workflow (in JSON or TS), then run it as a defined pipeline instead of ad-hoc glue. For a RevOps or growth engineer, that is the difference between a clever prompt and an operable process you can version and re-run. The honest read: it is two days old, beta, and has 20 stars — the idea is well-placed, but the plumbing is unproven, so pilot it on something non-critical before you build a GTM process on top of it.