Documentation
Build on MZGrid
Everything you need to deploy workloads, operate a provider node and integrate with the protocol — from a five-minute quickstart to every contract event.
Quickstart
Deploy Llama 3.1 8B in minutes
Run a provider
Install, benchmark, start earning
API reference
Every endpoint and response shape
Getting started
- IntroductionWhat MZGrid is, who uses it, and how a GPU job moves from a funded escrow to a settled payment on Base.
- QuickstartDeploy an OpenAI-compatible Llama 3.1 8B Instruct endpoint on MZGrid in a few minutes, from the website or from the TypeScript SDK.
- Core conceptsOffers, SKUs, templates, escrow, settlement, provider lock, reputation, job caps, regions and provider tiers explained.
Customers
- Deploying workloadsThe full customer flow: choose an offer or let the scheduler pick, configure a template, fund the escrow on Base and follow the job to completion.
- Billing & escrowHow deposits, per-second billing, the 7% protocol fee, refunds, cancellation, timeout reclaim and slashing compensation work, with worked MZG examples.
- TemplatesEvery workload template — what it runs, minimum VRAM, recommended GPUs and an example request.
- GPU catalogueThe supported GPU models, their reference prices in MZG per GPU-hour, how live prices relate to them, and how to choose a GPU for your workload.
Providers
- Installing the provider agentRequirements, installation, the mzgrid-provider CLI, running the agent as a systemd service, and how jobs are isolated on your machine.
- Benchmark & verificationHow MZGrid verifies provider hardware and work: registration benchmarks, usage reports and canary spot checks.
- PricingSetting your price per SKU, how live prices relate to reference prices, and how the scheduler ranks offers.
- Lock & exitHow 20% of every payout becomes your collateral, how it raises your job cap, and how to exit and withdraw it after 10 days.
- SlashingThe three slashing reasons, the exact amounts, how slashed MZG is split between the client and the treasury, and worked examples.
- ReputationHow the 0–100 reliability score is computed from completion rate, uptime, benchmark consistency and tenure, and how it affects ranking.
API reference
- Overview & authenticationBase URLs, conventions for amounts and timestamps, wallet sign-in, API keys, provider request signing, errors and rate limits.
- Public endpointsUnauthenticated endpoints for health, the GPU and template catalogue, marketplace offers and live network statistics.
- Jobs & accountCreate, fund, monitor, read logs from and stop jobs, and read your account balances. Includes the Job object and status reference.
- API keysCreate, list and revoke long-lived mzg_live_ API keys for servers and scripts.
- Provider endpointsEndpoints used by the provider agent: registration, benchmarks, heartbeats, job polling, status, usage, completion, account and pricing.
SDK
- TypeScript SDKInstall @mzgrid/sdk, create a client with an API key or wallet, and browse offers, create and fund jobs, wait for endpoints, read logs and stop jobs.
- Using the API from PythonCall the MZGrid REST API from Python with requests, fund jobs with web3.py, and use the job's endpoint with the openai client.
Smart contracts
- Overview & addressesThe three MZGrid contracts on Base, their deployed addresses, roles, the admin multisig and the trust model.
- MZGridEscrowConstants, functions, rules and events of the escrow contract, with a viem example of createJobWithPermit.
- ProviderRegistryProvider status, locked balances, job caps, exit and withdrawal, slashing reasons, functions and events.
- MZGTokenThe MZG ERC-20 token: 18 decimals, EIP-2612 permit, fixed supply of 1,000,000,000, no mint and no owner, plus the allocation and vesting wallets.