Customer support
Power agents that resolve tickets, answer questions, and support customers across every time zone
AI powered by every device
Join a supernetwork of intelligence powered by devices like yours. Pay less for models or earn by contributing compute.
Model/auth provider
Aquaduck (API Key)
$ monitor my inbox for renewals and draft a Friday brief
✓Thinking...
✓Analyzing sources...
✓Creating background session...
✓Scheduling inbox monitor...
✓Validating results...
✓Done
Session usage: 1,458,825 tokensAquaduck cost: $0.37 | Alternative provider: $3.34
Cost-efficient inference for agents like
Claude CoworkCodexOpenClawPiOpenCodeHermesLangChainOpenHandsClaude CoworkCodexOpenClawPiOpenCodeHermesLangChainOpenHands
Aquaduck unlocks crowd computing for intelligence, layering in routing and optimizations for cost-efficient inference.
Contributors earn by helping power the supernetwork on their computers.
Inference API
Cost-efficient inference designed for how agents operate.
Get API accessHalf the price of other providers. Cost optimizations built in at the gateway for added gains.
Available and reliable anywhere in the world. Inference at the edge.
Access to best-in-class open models that rival frontier on performance.
One-line switch to start using Aquaduck for model serving and inference for any agent. OpenAI API compatible.
Earn with Aquaduck
Earn money from your spare compute
Start earningDownload the Aquaduck desktop app and start earning in one click.
Earn in your local currency and withdraw earnings through Stripe.
Compatible with MacOS, Linux, and Windows laptops and personal computers. Support for additional devices is on the way.
Automatically earn based on what your computer GPU and RAM can handle. Share raw compute only. Nothing ever touches your files.
Built for agents that do more
From coding and research to SEO, browser automation, and multi-agent workflows, Aquaduck is designed for long-horizon and long-running workloads.
News briefings
Your agents can monitor sources, filter signal from noise, and deliver briefs on what matters
Content operations
Get your agents to draft, refine, and scale content across your channels in the background
Research pipelines
Run agents that gather sources, synthesize findings, and keep research moving
Browser automation
Lower costs for browser agents that navigate the web and complete multi-step tasks on your behalf
Agent orchestration
Coordinate specialized agents that hand off work and operate in parallel at scale
Benchmarks and metrics
Tokens per second
~60-65 tps. Higher is better.
Time to first token
~200-300 ms. Lower is better.
Cost per token
~50% compared with providers. Lower is better.