AssetLayerTESTNET

Send your agent. 
It can take it from here. 

One MCP endpoint. Your agent finds deals, reads data rooms, and invests — inside caps its owner signed on-chain.

AGENT SESSION● connected
# any MCP-capable agent
$ mcp connect assetlayer.org/api/mcp
→ delegation: valid · owner 0x56…f5A6 · claim #4
→ budget_left: $17,500 · per_deal: $5,000 · classes: RE, credit
# tools
list_deals · search_documents · read_document · create_proposal
→ read_documents(deal_7) · 9 files · valuation, leases, SPV terms
→ contribute(deal_7, $2,500) ✓ tx 0x6a…f0 · books to owner
Try it now

Spawn an agent. Watch it invest.

Two minutes, no code. A throwaway agent wallet is generated in your browser (the key never leaves it), your signature authorizes it, and the console below runs the same MCP calls a real agent makes — ending with an investment proposal only you can approve. Testnet funds, real rails.

01
Owner signs once
A verified human sets your budget, caps, and asset classes. One transaction. One registry claim.
02
You connect
Point any MCP agent at the endpoint. Deal terms are structured data. Documents are chunked for retrieval.
03
Everything books to the owner
Shares, refunds, income — everything accrues to the human. The contract enforces the caps, not a policy doc.

Limits the contract enforces.

Every call is checked against the delegation claim on-chain. Out of budget, wrong class, revoked — the transaction reverts.

Per-deal cap · total budget
Allowed asset classes
Instant revocation by the owner
Beneficial ownership: always the human
DELEGATION CLAIMagent 0x91aB…44Ceowner 0x56…f5A6caps enforced on-chain ✓
Build on it

Everything is documented.

The whole platform is an API. Every deal, document, and compliance check your agent can reach through MCP is also plain HTTP — same data, same rules.

For agents
MCP connect command, tool list, and the propose-and-approve flow.
Read →
Quickstart
From zero to your first authenticated call in a few minutes.
Read →
API reference
Deals, compliance, sessions, delegations, proposals, data rooms.
Read →
Contracts
Deployed addresses, ABIs, and the on-chain write surface.
Read →
POST /api/mcp
Read the docs →Browse deals

Live today: the playground above runs the full flow — delegation, agent session, MCP tools, propose-and-approve — and hands you the exact command to plug your own agent in (see For agents in the docs). On-chain spend caps (ADR-0009) are implemented in the v2.1 vault — grant your agent a contract-enforced budget and it executes within it; audited before mainnet.