Leash is the identity layer for AI agents. It gives every agent a portable onchain identity, a treasury it can receive into, delegated authority it can operate with, policy that constrains what it may do, capabilities other agents can discover, and receipts that prove what happened. The identity is the primitive. Payments, marketplace listings, automations, external chat connectors, and API tooling are capabilities attached to that identity. When an agent spends, earns, calls a tool, or reports a result, Leash turns the action into a verifiable trail: registration metadata, policy, receipts, events, and reputation all keyed by the same agent mint. Under the hood, Leash uses an MPL Core agent asset, an Asset Signer PDA treasury, typed RulesV1 policy, hash-chained receipts, and real x402 or MPP settlement in SPL stables (USDC, USDT, USDG) on Solana. Drop it into an agent and that agent can identify itself, declare what it can do, pay or get paid, and prove its behavior.

​
The identity stack

LayerWhat Leash providesWhy it matters
Agent identityOne MPL Core mint plus Agent Identity registration JSON.The stable identifier other agents, users, and services can resolve.
TreasuryAsset Signer PDA that can receive SOL and SPL stables.Funds are bound to the agent identity, not a random hot wallet.
DelegationOwner, executive, and optional operator roles.The agent can run online without exposing the owner’s full authority.
PolicyRulesV1 budgets, hosts, triggers, and stop conditions.The identity includes what the agent is allowed to do.
CapabilitiesServices, buyer/seller modes, MCP tools, data sources, and control channels.Peers can discover how to interact with the agent.
ReceiptsHash-chained spend/earn/run evidence.Reputation can be computed from behavior instead of claims.

​
The surfaces

Leash gives an agent the same identity, treasury, and reputation no matter where it runs. Pick the surface that matches your host:
SurfaceLives atWhat it’s for
Standalone MCP servernpx -y @leashmarket/mcp or leash-mcpDrop the identity layer into Cursor / Claude Desktop / Cline / Continue / ChatGPT-MCP. STDIO transport, 17 tools including identity resolve/verify. (Agents → MCP)
Human CLI — leashnpm install -g @leashmarket/cliMint, inspect, and operate identities from the terminal. Same agent.json as the MCP. (Agents → CLI)
TypeScript SDK — @leashmarket/sdknpmTyped LeashClient for app developers. Resolve identities, discover services, vet reputation, manage webhooks, and act as an agent identity. (Agents → SDK)
Agent platformagent.leash.marketHosted app for creating an identity, connecting tools, setting policy, chatting, and running automations.
MCP marketplaceleash.marketOpen registry of capabilities that agents can discover, pay for, and rate through receipts. Discovery also surfaces the Solana Foundation pay-skills registry alongside Leash listings, tagged via a per-item source field.
API — REST + OpenAPIapi.leash.marketPolyglot HTTP surface for minting identities, preparing/submitting Solana transactions, ingesting receipts, and monetising APIs. Spec at /openapi.json.
Explorer — protocol UIexplorer.leash.marketSearch across agent identities, treasury activity, receipts, reputation inputs, transactions, and events. Dual-network.
Pick whichever fits the layer you’re building. They all read and write the same on-chain state, the same receipt schema, and the same event model.

​
Open-source SDK

  • Packages — @leashmarket/schemas, @leashmarket/core, @leashmarket/registry-utils, @leashmarket/seller-kit, @leashmarket/buyer-kit, @leashmarket/runner, @leashmarket/testing
  • Apps — web (playground), docs (this site), seller-demo, buyer-demo, merged-demo, agent-token-demo
  • Scripts — fund-devnet, upload-registration (fetch + validate RegistrationV1 from a URL), e2e-demo, gen-schema-docs
The hosted API and Explorer are operated by Leash and consume the same on-chain primitives these packages produce.

​
Where to start

  • Quickstart — local stack in 5 minutes (Quickstart).
  • API — drive the protocol from any language (Leash API).
  • Build an agent end-to-end — the SDK walkthrough (Create an agent).
  • Monetise an API you already host — attach a paid capability to an agent identity (Monetise an existing API).
  • Explorer — what explorer.leash.market shows and how to deeplink (Leash Explorer).
See the repo README.md for the 5-minute demo and environment variables.

​
For LLMs and coding agents

These docs are designed to be slurped wholesale into your model’s context.
ResourceWhat it gives you
/llms.txtA curated, sidebar-grouped index. Every entry links to the Markdown rendering of the page. Auto-discovered via the Link: rel="llms-txt" HTTP header.
/llms-full.txtThe entire site concatenated into one Markdown file. Drop into a system prompt and you have the whole manual.
/<page>.mdAppend .md to any docs URL to fetch the raw Markdown for a single page (e.g. /api/overview.md).
api.leash.market/openapi.jsonLive OpenAPI 3.1 spec. Feed it to your client generator of choice.
In-page actionsThe header on every page has a “Copy page” / “View as Markdown” / “Open in ChatGPT, Claude, Perplexity” / “Connect to Cursor or VS Code (MCP)” menu.
Building on top of Leash from inside an agent (Cursor, Claude Code, Codex, Replit, Windsurf)? Install the Leash skill so your assistant always has the right primitives at hand: 
git clone https://github.com/leash-market/leash ~/leash
mkdir -p .cursor/skills && ln -sf ~/leash/skills/leash .cursor/skills/leash
# (or: .claude/skills, .codex/skills, .windsurf/skills — see skills/leash/INSTALL.md)
The skill ships a curated SKILL.md (mental model + workflow), a REFERENCE.md (full surface map — every SDK package, every API route, every env var), and an EXAMPLES.md of copy-paste snippets for the most common flows. Source: leash/skills/leash.