Cline now runs on Vercel AI Gateway
By Steven Van ·
Cline's coding agent moved its model traffic onto Vercel's AI Gateway, cutting P99 latency and error rates in a week-long A/B test.
Vercel AI Gateway is now the routing layer behind Cline, the open-source coding agent used by more than 1 million developers across 4 million installations. Requests from Cline's IDE integration now pass through Vercel's global network, hitting one of more than a hundred points of presence before travelling over Vercel's private backbone to the nearest region, which the companies say brings sub-20ms overhead for interacting with models through Cline.
Cline ran a week of live A/B testing against its previous AI router. The announcement reports:
- P99 streaming latency improved by 10 to 14 percent across Cline's most-used models
- API error rates dropped by 43.8 percent
- Grok Code Fast 1 saw 9 percent faster P95 latency and 13.7 percent faster P99 latency
- Minimax M2 delivered 14.4 percent faster P99 latency, along with cost reductions from more efficient routing
Cline is now the largest opted-in application on the AI Gateway Leaderboards, and became the top app by usage as soon as it switched its traffic to the Gateway. On pricing, Vercel says there is no markup on direct requests and no markup on Bring Your Own Key usage: developers pay what the model provider charges. The Gateway also gives Cline's team telemetry, health checks and per-request insight for diagnosing issues and planning capacity.