Nitrograph

Nitrograph is the trusted interface for every API: your agent gets one key and one balance, discovers the best-ranked service for any job, and only pays for validated answers. Failed calls are refunded; every charge carries a signed receipt.

Start here

Quickstart - key, connect, first paid call in five minutes. Works with Claude Code, Cursor, Windsurf, Claude Desktop, Codex, and any MCP client.
MCP tools - the tools your agent sees and when to use them.
HTTP API - invoke, credits, keys, receipts, raw.
Agent harness - the TypeScript SDK.
Client setup - per-editor install instructions.

Discovery is free and keyless. Paid invocations use your API key (ng_live_…, created on the dashboard) against prepaid credits - see pricing.

The platform underneath

The commerce layer runs today on established settlement infrastructure. Nitrograph also develops its own base network - a separate, clearly-gated project with its own status:

Layer	What it is	Status
Applications	Discovery + validated commerce (this site)	live
Network	Nitrograph's own base ledger	certified base v0.1.0 · not live
Protocol	Agent-native identity/coordination layers	in design
Evidence	Independent audit + certification policies	published

Tracking the protocol? The agent-native layers are specified as they are built, with exact status labels.

Status vocabulary

Every capability in these docs carries one of these labels:

certified base: implemented, independently reviewed, part of the accepted v0.1.0 baseline;
active hardening: implemented, undergoing audit remediation and network qualification;
in development: being built now; interfaces may change;
planned: approved roadmap, not yet started.

If a page doesn't say it's live, it isn't.

For agents

If you are an agent reading this, jump to /llms.txt for a machine-oriented index, or /llms-full.txt for the full text reference. The status vocabulary above applies to machine surfaces identically: do not treat protocol-layer capabilities as callable.