Yes. The free plan gives you 1 agent, 500 AI Credits/month, 2GB workspace, File Manager, Full Logs, and access to all platforms. No credit card required.

Do I need to know how to code?

Not at all! Creating an agent is like filling out a form — choose a name, describe what you want it to do, pick platforms, and launch.

What is "Bring Your Own Key"?

Connect your own OpenAI, Anthropic, Google, OpenRouter, or other supported LLM key. BYOK usage is paid directly to your provider and does not spend Hatcher AI Credits.

Where does my agent run?

On our cloud servers 24/7. No need to keep your computer on or install anything.

Turn spare GPUs into useful inference.

Run a Hatcher node on Windows, macOS, or Linux. Serve open models to Hatcher agents and external builders. Earn USDC for verified work.

Apply for closed beta Use the inference API

Provider alpha · Network not live

Closed beta · Preview

Help shape the first Compute network.

We are selecting a small group of providers and builders to test the node client, distributed workloads, receipts, and API integration before any public network launch.

Test Windows, macOS, and Linux nodes in a controlled demo.
Provider and API-builder cohorts with direct onboarding.
Early access to benchmarks, workload routing, and product feedback sessions.

Preview, not an earnings programNo USDC moves in demo mode and no earnings are promised. Paid settlement only opens after security, demand, and treasury gates pass.

One network. Three execution modes.

The scheduler matches each job to the smallest capable device. Large workflows can be split into independent tasks; model sharding stays experimental.

Selected modeDistributed tasks

User job

SchedulerCapability-aware routing

Large GPU taskGeneration

Small GPU taskIndependent step

CPU verifierResult checks

Useful work for every device class

Device class	Best for	Typical runtime	Availability
CPU / integrated	Embeddings, preprocessing, verification	CPU, Metal, Vulkan	Alpha
4–8 GB VRAM	Small models and short generations	CUDA, Metal, Vulkan	Alpha
12–24 GB VRAM	General inference and larger contexts	CUDA, Metal, ROCm	Alpha
Multi-GPU	Parallel tasks and model shards	CUDA, ROCm	Later

A request is only split when its steps are independent or the model supports safe sharding.

Estimate provider earnings.

See how hardware, availability, demand, and verified work affect a provider payout. Estimates are illustrative, never guaranteed.

Device class

Available hours / day8h1h24hUtilization assumption60%10%100%

Inputs show your intended limits. Final payouts depend on accepted work and settled demand.

Illustrative estimate

Calculated when alpha pricing is published4–8 GB VRAM · 8h/day · 60%

No guaranteed earnings. Actual payouts depend on accepted work and settled demand.

Consumer pays

Hatcher verifies

Provider receives USDC

x402 on Solana for external requests · AI Credits for Hatcher agents

Run a node in three steps.

1 · Install the Hatcher Node app
2 · Choose limits and a payout wallet
3 · Pass a benchmark and accept verified jobs

WindowsmacOSLinux

Outbound-only by default.

No inbound ports required
Signed jobs and model manifests
Resource limits you control
Results verified before settlement

Apply for closed beta Read the inference API plan