Pollinate

Pollinate is best understood as a data plane for agents rather than a data pipeline for analysts. The distinction matters: Fivetran and Airbyte move data toward warehouses and BI dashboards, while Pollinate's stated job is to collect, own, and route data to multiple agents with agent-aware formatting and schema design, so you are not re-serializing outputs for every model you plug in. Two surfaces tell you who this is for. The open-source npm SDK, khotan-data, lets a data engineer instrument an existing stack directly in code, and the SSH terminal at ssh.pollinate.tech means the vendor is comfortable selling to people who live in a shell. Neither is a drag-and-drop integration UI, and nothing in the public material suggests you should expect one. The genuine strength is the routing model. If you feed several agents from many sources — warehouse inventory signals into a demand forecasting agent, supplier shipment signals into a procurement assistant, live logistics alerts into a custom agent — the value is not the ingestion itself but the machine-shaped formatting and the fact that one collection layer serves all consumers. That is a narrower and more useful promise than a generic ELT tool makes. The risks are equally concrete. The company is YC W26, so references, security certifications, and production SLAs are not things the public surface establishes. The vendor page itself now leads with a core announcement that Pollinate is now Khotan and that Khotan builds bespoke procurement software around teams' rules, approvals, and exceptions. That is a meaningful pivot signal if your evaluation is based on the original agent-data-plane positioning — the brand you signed up for and the product roadmap may no longer point the same direction. Public documentation reachable this run is thin, and no pricing page or integrations page was reachable, so treat any assumption about tiers, free plans, or a connector catalog as unverified. Where it fits: AI-native startups and data engineering teams whose bottleneck is feeding scattered signals into agents, who are willing to run a scoped pilot, and who value a code-first SDK over a visual builder. Where it does not: teams that need proven SLAs and certifications before signing, non-technical users wanting a no-code integrator, and anyone looking for a finished analytics dashboard rather than infrastructure.