Agent Payments · Early Access

Financial infrastructure for AI agents.

Connect your agent to Points. Give it permission to discover opportunities, prepare payments, and operate within financial policies you control.

Payment tooling today creates drafts for operator review. Approve does not execute. Execution: not enabled. Capabilities roll out progressively in Early Access.

Connect an Agent Create Agent Account

Operator in the loop

Agent Connections keep identity, scopes, and approvals with the human operator — not an open autonomous wallet.

Back to AI Agents

How it works

Four steps from connect to controlled operation.

Step 1

Connect an agent

Link ChatGPT, Claude, Gemini, Grok, Perplexity, or a custom MCP client through Agent Connections.

Step 2

Grant scoped permissions

Choose read-only discovery, recommendations, or limited autonomy that stages work inside your policies.

Step 3

Discover and prepare

Your agent can surface opportunities and prepare payment drafts. Drafts require operator approval before anything moves.

Step 4

Operate under control

Spending limits, approval thresholds, and audit trails stay with you. Points does not claim unsupervised payment execution today.

Illustrative product example

Not live transaction data

What an agent payment draft can look like.

This card explains the product shape. It is not a live balance, quote, or executed payment. Points Pay capabilities remain non-executable until an authorized live method is available.

Agent

Ops Assistant

Draft amount

$42.00 USD

Status

Awaiting operator approval

Drafts can be created over MCP or Agent Connections with policy context attached.
Approval stays with the operator. No tool in the current MCP surface executes money movement.
Fees and settlement rails are disclosed only when a live delivery method is authorized — they are omitted here because they are not generally available yet.

Permission model

Clear modes. Honest boundaries.

Limited Autonomy means prepare and stage within policies plus operator approval — not unsupervised spending.

READ ONLY

Balances, offers, and transaction summaries — discovery without proposing money movement.

RECOMMEND

Suggestions and optimization hints the operator can accept or ignore. A recommendation alone never moves money.

LIMITED AUTONOMY

Prepare and stage payment drafts within policies you set. Operator approval is required. This is not unsupervised spend.

Use cases

Built for agent-shaped work.

API and compute spend

Let an agent draft purchases for tools and inference while you keep approval and limits.

Supplier and workflow payouts

Stage recurring operating payments from a named agent workspace with an auditable trail.

Rewards-aware commerce

Discover offers and PTS reward context alongside payment preparation — rewards are a separate economic layer.

PTS · rewards / economic layer

POINTS product rails and PTS stay distinct.

PTS is the network rewards and economic-coordination layer. It is not required to connect an agent, and this page does not make price, return, or investment claims.

Explore PTS for Agents PTS transparency

Built for control

Policies, approvals, and auditability first.

Scoped OAuth and Agent Connections — secrets never appear in the public catalog.
Approval-required defaults for connections that can stage financial actions.
Operator workspace for supervision after Early Access enablement.

Continue in the workspace

Authenticated operators manage connections under Agent Connections. Public pages stay informational.

Connect an Agent Review policy shell

Developers

Points MCP for builders.

Streamable HTTP MCP documents read-only tools and payment-draft proposals. Financial sources are sandbox-labeled where applicable; drafts never execute payments.

Open builder docs

Genesis cohort

Join the Genesis Cohort.

Early operators helping shape agent financial infrastructure. Same secure signup — attributed so we can measure cohort conversion.

Join the Genesis Cohort Create Agent Account