Practical overview
When it fits
Use Hypha when an LLM agent requires a structured runtime around reasoning and tools, especially where domain configuration, capability governance, durable execution, recovery, custom FSM control or validated-work reuse matter.
Who it is for
- TypeScript developers building production-oriented LLM agent applications.
- Teams that need governed and recoverable agent execution.
- Builders creating reusable domain-specific agents with shared runtime components.
Documented uses
- Define domain-specific tasks, workflows, capabilities, memory, policies, evaluations and output contracts through DomainPacks.
- Control agent execution with finite-state-machine flows, events and checkpoints.
- Support approval, continuation, recovery, replay and audit around agent work.
- Reuse validated reasoning, tool, execution, memory and inference work through typed cache layers.
- Compose custom human-in-the-loop or domain-agent workflows.
Requirements & limits
The repository identifies Hypha as open source under the Apache-2.0 license. It is presented as a TypeScript framework for CLI, server and self-hosted use. No pricing or account requirement is documented here.
- Domain behavior is separated from the shared runtime through DomainPack.
- The Production Harness adds FSM control, events, checkpoints, policy and approval, recovery, replay and audit.
- Cache layers have different validity and invalidation boundaries for responses, reasoning, work, tools, memory and prompt prefixes.
- The repository and linked guide identify custom FSM control and package-level API coverage.
- The supplied material does not include a complete installation, configuration or execution sequence.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 3 October 2026
- Original website title
- CodeSoul-co/Hypha · Official homepage title · retained in its source language

