Tested on NVIDIA DGX Spark and Brev

Press Use. Your agent runs it.

YAAD is an app store where the apps are finished work: expert Stacks and skills that your own AI agent runs on your GPU, or on ours free to start. Every run comes back with a receipt.

Sign in with your email. No card, no hardware needed to try it.

The YAAD client: Discover, with a featured Stack ready to Use
1,230+GPU jobs in our NVIDIA stress test; 96% succeeded while we crashed machines on purpose
0 steps lostin 11 of 11 live moves between a DGX Spark and cloud H100s
55 skillsfor 8 kinds of work, picked by your agent in one call
0 bytessent to the internet from inside a run. Your data stays put.

Measured on YAAD in November 2026.

How it works

Your agent is the interface. YAAD does the rest.

Link the coding agent you already use. It finds the right Stack, skill or capability, runs it on the cheapest machine that fits, and shows you what happened. You approve anything that costs money.

1

Connect your agent

Claude Code, Codex, ChatGPT, Cursor, Gemini and more sign in with one YAAD ID.

2

Press Use, or just ask

Pick a Stack someone already ran, or tell your agent the job. It pulls the right skill and capabilities by itself.

3

It runs where you choose

Your DGX Spark, NVIDIA Brev or our free GB10. If a machine goes down, the job carries on from its last checkpoint.

Connect your agents: Claude Code, Claude, Codex, ChatGPT, Grok, Gemini CLI, DeepSeek, Cursor
Skills

Expert know-how your agent picks up in one call.

We went through 10,355 public skill collections and merged the best of about 1,170 skills into 55, one per job: rewritten, tested, every source credited. Your agent calls use_skill and follows the right one.

ResearchersStudentsAI builders AccountantsMarketersProduct managers Builders3D / CAD engineers
Skills in the YAAD catalog: ad creative, agent and tool design, AP and AR invoice control, build plan
Devices

Every machine you use, on one plane.

Add your own hardware with one command, or rent cloud GPUs. YAAD places each job on the first machine that fits and moves running work between them with its checkpoint: in our tests, a median 22-second handover with no steps redone.

NVIDIA DGX Spark
Your DGX SparkNVIDIA GB10, 128 GB unified memory$0 on hardware you own
NVIDIA H100
NVIDIA Brev H100For heavy training and long contextBy the minute, inside your budget
Server rack
Your servers and labAny Linux machine with Docker$0 on hardware you own
Storage
Your filesSent straight to your machines, never through a cloud copyPrivate by default
Everything on YAAD

See it work. Press Use. Your agents run it.

  • Stacks with a record. Each one shows where it ran, how long it took and what it cost, before you press Use.
  • 53 tested capabilities. Sealed functions any agent can hire for one job: clean a table, read an A/B test, plan GPU memory, scan for secrets.
  • One key for every model. Your YAAD ID reaches model and compute providers, routed to the best price for each call.
Everything on YAAD: Stacks such as QAOA on CUDA-Q, CPU test suite, GPU eval run, LoRA fine-tune
Community

Better with friends.

Follow the creators whose Stacks you use, add friends and see what they are running, read reviews written only by people who actually ran a Stack, and earn achievements as you go.

A home that tells you what happened

What finished while you were away, what needs your OK, who used your Stacks.

Friends and creators

See who is online and what they are running. Their shares and runs show in your feed.

Reviews from real runs

Only people who ran a Stack can review it, so the record and the reviews agree.

Privacy

Your data stays yours. By design.

Sealed runs

Jobs run with the network off. Models and data are fetched before the run starts; then nothing leaves.

A receipt for every run

Which machine, which image, how long, what it cost. Your agent and your team see the same record.

Keys stay in YAAD

Connect a provider once. Agents use it through your YAAD ID and never see the key.

Pricing

Pay for what runs. Nothing else.

Start free10 runs a day

YAAD's capabilities on our GB10, up to 5 minutes each. No hardware needed.

Your own hardware$0

Your DGX Spark, servers or lab machines. Run as much as they hold.

Cloud GPUsBy the minute

NVIDIA Brev and others, inside the budget you set. You see the price before it starts.

SkillsFree

All 55, for any agent you connect.

Teams, labs and hackathons

Bringing hardware or a team?

Anyone can start now with just an email. To connect a DGX Spark or a lab's servers, or to run a hackathon on shared GPUs, tell us and we'll help you set it up.

Get started free

We only use your email to reply.