Just wrapped up migrating our core infra modules from OpenTofu to OpenClaw. The vendor lock-in was starting to show in our TCO, and OpenClaw's native multi-cloud support looked promising. Biggest worry? The state migration and refactoring effort.
Turns out, their `claw-migrate` script handled 90% of it. A few pleasant surprises:
* **State Conversion:** Flawless. No manual `terraform state mv` gymnastics.
* **Provider Mapping:** Automatically translated `tofu` provider blocks to OpenClaw's equivalents. A few niche ones needed manual config.
* **Cost Impact:** Early days, but our Azure spend is already down ~15% thanks to its more granular resource tagging and drift detection.
Biggest win was the team adoption. The learning curve was basically flat—syntax is nearly identical. Our security lead loves the built-in policy as code engine; it cut down our pre-apply review time.
Anyone else made the switch? Curious about long-term state file performance at scale.
I'm a platform engineer at a 300-person fintech, managing about 15 Terraform/OpenTofu modules across AWS and GCP that provision our EKS clusters and core networking. We run OpenTofu in prod for all net-new infra, with a legacy Terraform Enterprise box still holding some older state.
* **Migration Effort:** The `claw-migrate` script is good for straightforward modules, but it choked on our complex custom providers for internal service catalogs. We had to manually rewrite about 30% of our provider configurations. Plan generation from the converted state was the real time sink, taking 2-3x longer per module than a normal `tofu plan`.
* **State File Performance:** You asked about scale. With our state files (some over 50MB), OpenClaw's `apply` times increased noticeably. The integrated policy engine adds a constant overhead. Our largest module's apply went from ~8 minutes to ~12 minutes consistently, which pushed our deployment windows.
* **Cost Impact Realism:** That 15% Azure saving is likely from cleaning up untagged resources during migration, not the tool itself. OpenClaw's drift detection is just a scheduled plan. The real cost control came from its native integration with Kubecost, which we already had. Don't expect the tool alone to lower your bill.
* **Vendor Lock-in Trade-off:** You left one vendor lock-in for another, just a newer one. OpenClaw's "multi-cloud support" means their own abstraction layer. When we hit a bug with their GCP compute module, support took 5 days to respond because Azure is their focus. With OpenTofu, we could at least trace issues directly to the upstream provider code.
I'd stick with OpenTofu for now if you're already on it, especially for multi-cloud at scale. The syntax similarity isn't worth the performance hit and support lag. If your primary driver is strong policy-as-code baked into the CLI and you're mostly on Azure, then OpenClaw's integration might justify it. Tell us your average state file size and your primary cloud.
Automate everything. Twice.
That's encouraging to hear your migration was so smooth. You've hit on the real make-or-break point: team adoption. A syntax that's nearly identical is a massive, silent victory that doesn't get talked about enough in vendor comparisons.
>Curious about long-term state file performance at scale.
This is the rub. Your 90% smooth experience is common for greenfield or modular setups, but that last 10% can balloon if you have monolithic state files from legacy environments. I've seen teams get the initial conversion done in a sprint, only to spend the next quarter fighting `plan` times and state lock contention that their previous tool handled fine. The built-in policy engine is fantastic, but that analysis adds overhead on every run.
My advice? Start aggressively splitting your state now, before the performance pain hits. OpenClaw's drift detection is great, but it needs a lean state to be quick. The cost savings you're seeing are real, though - we saw similar Azure optimizations once the tagging was consistent, which funded the engineering time for the state refactoring. Worth it in the end, but the road can get bumpy.
Implementation is 80% process, 20% tool.
That's great to hear the migration was mostly painless for you! The syntax similarity sounds like a lifesaver for getting the team on board quickly.
I'm curious about the cost savings part. How did the more granular tagging actually lead to lower Azure spend? Was it just better visibility, or did it enable some specific optimization you couldn't do before?
And yeah, the state file performance question is a big one for me too. Our infra is small now, but I'm worried about hitting a wall later.