New signups are closed. That's what you need to know first about Inferable—Xero acquired them. But it's not dead. MIT-licensed code means self-hosting works if you'll set that up.
The free plan? Two registered workflows. 1000 executions monthly. Max concurrency sits at 2. Paid starts at $10 per month—20 workflows and 5000 executions included. Extra runs cost $1 per thousand. Concurrency bumps to 10.
Here's what Inferable actually does: turns your existing codebase into conversational agents. Platform engineers drowning in fragmented internal tools can consolidate them without rewriting everything. You register workflows tapping into your APIs and data. Agents execute those workflows through natural language. Orchestration gets handled automatically so AI workloads don't fall apart mid-execution.
Workflow caching speeds up similar operations when you're running them repeatedly. You bring your own models—no vendor lock-in. Self-hosting means you control where everything runs.
Timing's unfortunate. This could've simplified internal tooling operations just as teams figure out how to deploy agents at scale. But the code's still there. You'll just spin up your own instance instead of using their hosted version. For developers comfortable with that trade-off? Still an option. Enterprise deployments worked with AWS, GCP, and Azure before acquisition happened.
