Areev Store
Get Started

For agents

Your agent can shop here.

Connect Claude, Cursor, ChatGPT or your own stack over MCP, A2A or REST. Discover outcomes, propose purchases, and call the work you already own — with a named human on every commit.

Read the integration guideAPI reference
  • MCP
  • A2A
  • REST + OpenAPI
  • Webhooks
~/.claude/mcp.json Copy
{
  "mcpServers": {
    "areev-store": {
      "url": "https://mcp.areev.ai/store",
      "headers": {
        "Authorization": "Bearer ${AREEV_KEY}"
      }
    }
  }
}

Four ways in

Pick the protocol your stack already speaks. They all reach the same catalogue and the same outcomes.

All interfaces read one registry
MCP

Model Context Protocol

For assistants and IDE agents. Tools, resources and prompts appear natively in the client.

mcp.areev.ai/store

ClaudeCursorAny MCP client
A2A

Agent2Agent

For agent-to-agent delegation. Send a task, stream progress, receive artifacts and evidence.

areev.ai/.well-known/agent-card.json

Agent CardTasksStreaming
REST

OpenAPI 3.1

For backends, scripts and anything else. Typed clients generated from the published description.

api.areev.ai/store/v1

OpenAPITyped SDKsIdempotent
Webhooks

Signed lifecycle events

For settlement. Every outcome state change is pushed to your endpoint, signed and replayable.

POST https://your-app/areev/events

SignedReplayableAt-least-once

One Store interface. One more per outcome you own.

The Store interface is how your agent finds and buys work. The moment an outcome is yours, it publishes its own interface — with the tools, inputs and evidence defined by that outcome's contract.

Step 1

Your agent

Whatever you already run. It speaks one of the four protocols.

Claude
Cursor
ChatGPT
Your own stack

Step 2 · one per account

The Store interface

Discover, compare and configure. Buying is proposed here, never committed here.

mcp.areev.ai/store

search_outcomesAutomatic
get_outcome_contractAutomatic
get_priceAutomatic
check_availabilityAutomatic
propose_purchaseYou approve
list_my_outcomesAutomatic

Step 3 · one per outcome

Outcomes you own

Each one publishes its own MCP server, Agent Card and REST base. Its tools come from its contract.

AI Clinic Receptionist

mcp.areev.ai/o/clinic-receptionist

$4·5 tools

NPHIES Claim Pre-Screen

mcp.areev.ai/o/nphies-pre-screen

SAR 3·4 tools

PE Deal Intake Agent

mcp.areev.ai/o/pe-deal-intake

$25·6 tools

Connect the assistant you already use

One key, one URL. No SDK required unless you want one.

Claude

Add the Store as a custom connector, or drop the config into Claude Code.

Settings → Connectors → Add

Cursor

Paste the same JSON into your global or project MCP settings.

Settings → MCP → Add server

ChatGPT

Add the Store as a connector, or build an action from the OpenAPI file.

Connectors → Add · or Actions

Your own stack

Generate a typed client from the published description, or call REST directly.

api.areev.ai/store/v1/openapi.json

Inside one outcome

What an outcome exposes the moment you own it. Nothing here is bespoke — the tools, skills and routes are generated from the outcome's contract, so every outcome in the catalogue works the same way.

AI Clinic Receptionist

Contract v3 · Areev Cloud Singapore · owned since March 2026

Active

$4/ verified booking

MCPTools

mcp.areev.ai/o/clinic-receptionist

  • check_availability
  • book_appointmentWRITES
  • reschedule_appointmentWRITES
  • cancel_appointmentWRITES
  • resolve_faq
A2ASkills

/o/clinic-receptionist/.well-known/

  • appointment.book
  • appointment.reschedule
  • faq.resolve
  • Streaming task updates
  • Evidence returned as an artifact
RESTRoutes & evidence

api.areev.ai/o/clinic-receptionist/v1

  • /appointmentsPOST
  • /outcomes/{id}GET
  • /outcomes/{id}/evidenceGET
  • /ledger?status=verifiedGET
  • /outcomes/{id}/disputePOST

What your agent may do on its own

The same approval boundary you set on the outcome page applies to the API. Calls that need a person return a pending decision — not an error — and your agent keeps its place in the queue.

ActionPolicy
Search the catalogue, read contracts and pricesAutomatic
Call an outcome you already ownAutomatic in contract
Read evidence and the outcome ledgerAutomatic
Buy an outcome or change a planHuman approval
Change deployment, region or keysHuman approval
Raise a dispute or accept a creditHuman approval
Act outside the outcome's contractNever

How work becomes billable

Every call writes to the Outcome Ledger. Your endpoint receives each state change, signed.

  • outcome.startedWork begins
  • outcome.completedEvidence attached
  • outcome.verifiedBillable
  • outcome.contestedUnder review
  • outcome.creditedNot charged

Your agent never decides what is billable. Deterministic checks settle the outcome; the agent can only read the result and explain it.

Four steps to your first outcome

Most teams are calling a purchased outcome the same afternoon.

1

Create a key

Account → Keys. Scope it to read-only or to the outcomes you choose.

AREEV_KEY=sk_live_…

2

Connect

Paste the config into Claude or Cursor, or point your A2A client at the Agent Card.

mcp.areev.ai/store

3

Buy an outcome

Your agent proposes; you approve in the Store or in your inbox.

propose_purchase(...)

4

Call it

The outcome's own tools appear in your agent the moment it goes active.

book_appointment(...)