For AI agents

Your AI agent can't drive. Our drivers can.

Connect your agent to the physical world. Use MCP or the REST API to check coverage, quote a route, book a driver and retrieve proof of delivery. Start in the sandbox, then move to live bookings when you are ready.

Agent setup Get API key

Your AI agentMCP

Uses a funded wallet and your spending limits.

Agent-native

It quotes.It books.It pays.It proves.

MCP tools for your agent: 43
quotes, no API key needed: Free
to a free sandbox key, no card: 60 s
funds are released only after proof of delivery: Held

Install

One line, then ask for a quote.

Add the hosted MCP server to Claude Code, or install it in Cursor with one click. Quotes and coverage work before you sign in.

Claude Code

claude mcp add --transport http rentadriver https://mcp.rentadriver.ai/mcp

Add to CursorOpens Cursor and installs the server.

Any MCP-compatible client works: Claude Desktop, Claude Code, Cursor, Windsurf, Cline, custom agents. Test with get_quote first — it needs no key.

Watch an agent book a real driver

From a chat message to a quote, a booked driver on the map and the proof photo.

From quote to handover

Know what happens at every step

01
get_quote
Free, no key. Price, ETA and a breakdown you can show the human before spending anything.
02
create_delivery
Funds are held from a prepaid wallet (card or USDC). dryRun first if you like.
03
dispatch
We offer the job to nearby drivers on shift. Track assignment and handle unavailable drivers in the same workflow.
04
track_delivery
Live status, driver position and ETA by polling, SSE or signed webhooks.
05
proof + confirm
Geo-stamped photos, signature or OTP. Confirm to pay the driver, or we auto-confirm in 24h.

See the full delivery cycle

The six tools most agents use

All 43 tools →

1check_coverage
Tells you whether an address or a lat/lng pair is inside a RentADriver service area, which city it maps to, and the nearest area with distance if it isn't covered..
2get_quote
Prices a delivery from one pickup to one or more drop-offs.
3create_delivery
Books a delivery from a quote_id or directly from pickup and drop-offs.
4track_delivery
Compact live view for polling: status, next_action, ETAs, driver first name and position, per-stop progress, last event and the SSE stream URL.
5get_proof_of_delivery
Signed photo URLs (valid one hour), signature image, recipient name, OTP result, coordinates and timestamps for the delivery and each stop..
6confirm_delivery
Confirms a delivered job, captures the held funds and releases the driver's pay.

Other clients and the full setup prompt

claude mcp add rentadriver -- npx -y rentadriver-mcp

# hosted server instead (sign in with OAuth when asked):
claude mcp add --transport http rentadriver https://mcp.rentadriver.ai/mcp

1. Paste this into your agent

Set up RentADriver so this agent can book real-world package deliveries.

Rule: only use RentADriver's own tools (check_coverage, get_quote, create_delivery, track_delivery, get_proof_of_delivery, confirm_delivery). If those tools are NOT connected in this chat, tell me they are missing and stop. Never quote a price from the website, a memory, or your own estimate as if it were a live quote, and never say something is booked unless a tool returned it.

1. Connect the tools:
   - Terminal agents (Claude Code, Cursor, VS Code, Windsurf, Codex, Gemini): run  npx -y rentadriver-mcp setup  (or add the MCP config from rentadriver.ai/for-agents).
   - Claude Desktop / claude.ai / ChatGPT — you cannot run npx here: add the remote connector  https://mcp.rentadriver.ai/mcp  in the app's connector settings and sign in when asked. In ChatGPT this is Settings → Connectors (developer mode) → add a custom connector.
2. If there is no API key yet, call the signup tool with my name and email — it returns a key once. Rehearse with a sandbox (rd_test_) key; live bookings need a funded wallet.
3. Test with check_coverage and get_quote (both free). Before booking, always call create_delivery with dryRun=true and show me the exact price the tool returns.
4. Book with create_delivery, then track_delivery until delivered, then get_proof_of_delivery and confirm_delivery.

3. Not on MCP? Use the REST API

# import the spec as tools
spec = requests.get("https://api.rentadriver.ai/.well-known/openapi.json").json()
# or hand-write the three that matter:
tools = [
  {"type": "function", "function": {"name": "get_quote", "description": "Price a delivery", "parameters": spec["components"]["schemas"]["QuoteRequest"]}},
  {"type": "function", "function": {"name": "create_delivery", "description": "Book a delivery (needs x-api-key)", "parameters": spec["components"]["schemas"]["CreateDeliveryRequest"]}},
  {"type": "function", "function": {"name": "track_delivery", "parameters": {"type": "object", "properties": {"delivery_id": {"type": "string"}}}}},
]

Design rules your agent can rely on

Price before money

get_quote is free and returns the full breakdown. create_delivery supports dryRun. Nothing moves until fund.

next_action everywhere

Every delivery response says what to do next: fund, wait, track, confirm, dispute, retry.

Idempotent creates

Send idempotency_key and a crashed retry returns the same delivery, not a second one.

Spending controls

Cap per-delivery and daily spend from the account; the API refuses anything above it.

Funds only move on outcomes

Held on create, released on cancel or no driver, captured on confirm, frozen on failure until a human reviews.

Proof is structured

Photos, signature, recipient name, OTP, geo and time — not a free-text claim.

What a completed record containsget_proof_of_delivery → proof

delivered

delivered_at: Timestamp of the handover, set by the app at the moment of completion
lat / lng: GPS position where the proof was captured
photos[]: Photo of the package at the door or in the recipient's hands
signature: Drawn on the driver's screen when a signature is required
recipient_name: Who took it, typed by the driver at the door

Proof, not a promise

Every handover leaves a record you can act on.

When a driver hands the package over, the app captures structured evidence and attaches it to the delivery. It is returned to your software as data, so an agent can check it, store it or dispute it without a human reading a screenshot.

You choose what counts as delivered

photoA picture of where it was left
signatureSigned on screen
recipient_nameThe recipient's name
otpA one-time code you gave the recipient
id_checkThe driver checks photo ID

Set proof_required when you book. If the driver cannot capture it, the delivery cannot be marked complete, and your funds are not captured.

See the proof API

Built for agents that spend real money

Guard rails your agent cannot talk its way past.

An autonomous agent booking physical deliveries is a new kind of risk. These limits are enforced by the platform, not by prompt, so a confused or compromised agent stays inside them.

Spending caps
Set a maximum per delivery and per day on the workspace. The API refuses anything above it, whatever the agent asks for.
Money moves on outcomes
Funds are held when a delivery is created, released if no driver is found or it is cancelled, and captured only when you confirm delivery.
Idempotent booking
Send an idempotency key and a crashed, retried agent gets the same delivery back, not a second driver.
Test keys, real API
Sandbox keys run the whole flow with simulated drivers and no money. Switch the key mode to go live; nothing else changes.
Price before commitment
Quotes are free, unauthenticated and return the full breakdown. Creates support dry-run. Nothing is charged until you fund.
Structured proof
Delivery evidence is data your agent can verify, not a free-text claim it has to trust.

Read the agent design rules

Built for your workflow

Not on MCP? Use the REST API

Connect through MCP or build with the REST API. Your agent can check coverage and budget, reserve funds, book a driver, track the delivery, review proof and confirm payment—all within your existing workflow.

Quotes and coverage checks are free and unauthenticated
Preview supported actions with dryRun before making a live request
Every response tells the agent its next_action
Spending caps per delivery and per day
Choose card or supported USDC funding for your delivery wallet

Agent setup OpenAPI spec

Set up RentADriver so this agent can book real-world package deliveries.

Rule: only use RentADriver's own tools (check_coverage, get_quote, create_delivery, track_delivery, get_proof_of_delivery, confirm_delivery). If those tools are NOT connected in this chat, tell me they are missing and stop. Never quote a price from the website, a memory, or your own estimate as if it were a live quote, and never say something is booked unless a tool returned it.

1. Connect the tools:
   - Terminal agents (Claude Code, Cursor, VS Code, Windsurf, Codex, Gemini): run  npx -y rentadriver-mcp setup  (or add the MCP config from rentadriver.ai/for-agents).
   - Claude Desktop / claude.ai / ChatGPT — you cannot run npx here: add the remote connector  https://mcp.rentadriver.ai/mcp  in the app's connector settings and sign in when asked. In ChatGPT this is Settings → Connectors (developer mode) → add a custom connector.
2. If there is no API key yet, call the signup tool with my name and email — it returns a key once. Rehearse with a sandbox (rd_test_) key; live bookings need a funded wallet.
3. Test with check_coverage and get_quote (both free). Before booking, always call create_delivery with dryRun=true and show me the exact price the tool returns.
4. Book with create_delivery, then track_delivery until delivered, then get_proof_of_delivery and confirm_delivery.

Connects to

Claude
Cursor
Windsurf
OpenAI
LangChain
CrewAI
Zapier
n8n
Shopify

MCP

One line, then ask for a quote.

Free sandbox key · no card

Get API key Agent setup

quotes	30 / min per IP
create delivery	120 / min per key
public tracking	120 / min
API keys	10 active
batch	25 per call

It quotes.It books.It pays.It proves.

One line, then ask for a quote.

Watch an agent book a real driver

Know what happens at every step

The six tools most agents use

Other clients and the full setup prompt

1. Paste this into your agent

3. Not on MCP? Use the REST API

Design rules your agent can rely on

Every handover leaves a record you can act on.

You choose what counts as delivered

Guard rails your agent cannot talk its way past.

Spending caps

Money moves on outcomes

Idempotent booking

Test keys, real API

Price before commitment

Structured proof

Not on MCP? Use the REST API

Connects to

One line, then ask for a quote.