Skip to content
Hightop docs header art
Hightop
API and Integrations

Build on Hightop#

If you already have agents in your own stack and need them to handle money safely, Hightop is the financial layer for that. Your agents stay where they are. Hightop gives them a bounded way to pay vendors, collect revenue, manage idle cash, and use shared capital through the API.

For the broader product model, see How Hightop Works. This page focuses on the practical integration pattern for builders.

What You Keep, What Hightop Takes Over#

You keep:

  • your agent logic
  • your prompts, workflows, and scheduling
  • your app, service, or agent runtime
  • your business-specific decision-making

Hightop takes over:

  • wallet and account infrastructure
  • policy enforcement for what each agent can and cannot do
  • payment paths for approved vendors, ad hoc payouts, and trusted destinations
  • shared capital management across payments, yield, borrowing, and rebalancing
  • human monitoring and revocation through the app

The Integration Model#

At a high level, the integration model looks like this:

your agent runtime
  -> Hightop API
  -> control layer
  -> smart contract wallet
  -> payment rails and onchain protocols

In practice, the loop is:

  1. You fund one Hightop account.
  2. You create one or more Hightop agents in the app. A Hightop agent is the control profile attached to your external agent, not the external agent runtime itself.
  3. For each one, you define its lane: permissions, limits, allowed assets, payment paths, approved destinations, and expiry.
  4. Your external agent connects to Hightop with that agent's API key.
  5. Your external agent sends requests through the API inside that lane.
  6. Hightop validates the request against the configured rules before funds move.
  7. If the request fits the lane, the wallet executes it. If not, the request is blocked.
  8. You monitor activity, alerts, and account state in the app.

The important design point is that your agents do not control money directly. They operate through the lane you configured.

Good First Integrations#

The best first integrations are narrow and operational.

Payments Agent#

Give one agent permission to pay a small set of approved vendors in USDC with tight per-transaction and monthly caps.

This is often the fastest path to production value: inference APIs, compute providers, data vendors, and recurring service costs.

Cash-Management Agent#

Give another agent permission to move idle balances into Earn, pull them back when cash is needed, and rebalance within a limited asset set.

This works well when you already have money flowing through Hightop and want a controlled treasury lane.

Revenue Collection Agent#

Let an agent receive usage-based payments or service revenue into the same account, then route those funds into the operating pool your other agents use.

Tightly Scoped Borrowing Agent#

If your workflows need working capital before revenue settles, you can give an agent narrow borrowing authority with clear asset and limit boundaries.

How to Start Without Over-Delegating#

The safest first rollout is usually the narrowest one.

Start narrower:

  • one account
  • one agent
  • one job
  • one payment path
  • one or two approved vendors
  • one asset, usually USDC
  • short expiry
  • tight transaction and period caps

Once that lane has proven itself, add the next one. This keeps the first rollout easier to reason about, monitor, and revoke if needed.

What to Expect#

Credentials#

Each external agent authenticates with its own API key. That key lets the agent act only inside the lane you configured for it. If a key is compromised, the attacker gets that lane, not broader authority. For the trust model behind that, see Security and Control.

Agent Lifecycle#

New agents do not always become usable immediately. The lifecycle is:

Created -> Waiting -> Active -> Expired or Removed

If an activation delay applies, a newly created agent can exist but still be unable to act until the waiting period ends.

Blocked Requests#

When a request falls outside the lane, Hightop blocks it before funds move. For the fuller request-outcome model and the common causes, see API Overview. For the debugging flow, see If an Agent Goes Off-Script.

Monitoring#

The app gives you visibility while agents operate. You can review account activity in the app, and Hightop surfaces real-time notifications when agents act, when limits are reached, when one-off payouts are pending review, and when critical changes are initiated.

Where to Go Next#

Previous

If an Agent Goes Off-Script

Next

API Overview