Tigris
An engineering blog making one argument: an agent's working state belongs in a forkable bucket
A side-by-side editorial comparison of Refine and Rivet — release velocity, themes, recent moves, and the top alternatives to consider.
| Feature | Refine | Rivet |
|---|---|---|
| Sector | DevOps | DevOps |
| Velocity score | 0.0 | 7.5 |
| Sparks · 30d | 0 | 1 |
| Top themes | react, internal-tools, admin-panels, access-control | agentos, coding-agents, webassembly, sandbox-alternative |
| Last editorial update | 2h ago | 1d ago |
| Website | Visit → | — |
Refine's v5 majors are out; what's shipping now is the bug tail, mostly from contributors
Refine is a React framework for building internal tools and admin panels, and this window is entirely patch releases against the already-shipped v5 core and v8 MUI packages. The fixes cluster in three places: URL state parsing under syncWithLocation, access-control metadata not reaching button hooks, and CLI binary resolution inside monorepo workspaces.
Rivet is betting coding agents don't need a VM — and other platforms are taking the bet.
Rivet's centre of gravity has moved from actors and serverless compute to agentOS, a WebAssembly and V8-isolate layer that gives a coding agent a Linux-shaped terminal, filesystem, networking, and dev servers without a sandbox VM behind it. Since the v0.2 release in June the pieces have arrived quickly: a package registry with 130µs installs, durable timezone-aware cron for actors, and a Rust rewrite of Secure Exec. The argument running through the writing is a cost one — sandboxes are heavy, and most agents only ever needed the operating system.
Refine is a React framework for building internal tools and admin panels, and this window is entirely patch releases against the already-shipped v5 core and v8 MUI packages. The fixes cluster in three places: URL state parsing under syncWithLocation, access-control metadata not reaching button hooks, and CLI binary resolution inside monorepo workspaces.
The pattern is a framework absorbing the edge cases that only appear once a major is in real projects. Deeply nested conditional filters were being mangled because a parser depth limit was too shallow; useDeleteButton was dropping the meta prop that custom access rules depend on; refine dev broke under workspace hoisting. Notably, most of these carry different contributor names, several first-time — the maintenance load is being spread across the community rather than driven by a single team push.
Filters, access control, and router adapters have each produced multiple fixes in a single window, so expect continued patch releases in those areas; nothing in these entries points to a new major or a change in framework direction.
Rivet's centre of gravity has moved from actors and serverless compute to agentOS, a WebAssembly and V8-isolate layer that gives a coding agent a Linux-shaped terminal, filesystem, networking, and dev servers without a sandbox VM behind it. Since the v0.2 release in June the pieces have arrived quickly: a package registry with 130µs installs, durable timezone-aware cron for actors, and a Rust rewrite of Secure Exec. The argument running through the writing is a cost one — sandboxes are heavy, and most agents only ever needed the operating system.
The pattern this month is distribution, not features. Vercel's Eve and Flue both adopting agentOS means Rivet is being consumed as a runtime layer inside other people's agent products rather than sold as a destination platform, and the registry gives that layer a way to ship dependencies. The latest technical post pushes the boundary further out, arguing the agent harness belongs in the customer's own backend with agentOS underneath it.
Expect more third-party agent platforms announced as agentOS adopters, and the registry to accrue first-party packages that make the runtime harder to substitute. The open question the entries do not answer is how agentOS is priced when it runs inside someone else's infrastructure.
Other DevOps products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either Refine or Rivet.
An engineering blog making one argument: an agent's working state belongs in a forkable bucket
Workato is packaging its agents into products — IT Support and EDI Genies launched a day apart
Lokalise is building the measurement layer around AI translation, not just more AI translation
Okta is documenting its way into agent identity, one Cross App Access guide at a time
HashiCorp is wrapping Terraform in policy and stacks for the agent-driven era.
GitHub is hardening the registry it owns while Copilot absorbs every new model.
See all Refine alternatives → · See all Rivet alternatives →
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
They serve adjacent needs but don't currently overlap on shipped themes. Rivet is currently shipping more aggressively (velocity 7.5 vs 0.0), with 1 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. Rivet is currently shipping more aggressively (velocity 7.5 vs 0.0), with 1 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other DevOps products to evaluate alongside.
Top Refine alternatives in DevOps are ranked by recent ship velocity. Browse the "Refine alternatives" section above for the current picks, or visit /alternatives/refine for the full list with editorial commentary on each.
Top Rivet alternatives in DevOps are ranked by recent ship velocity. Browse the "Rivet alternatives" section above for the current picks, or visit /alternatives/rivet for the full list with editorial commentary on each.