Product
Everything around the agent, built in.
Heighliner is infrastructure for running AI agent fleets, not the agents themselves. You bring the agent. Every fleet gets the same five controls, whatever runs inside it.
Fleets
The unit you deploy is a fleet.
A fleet is one to many agents running one harness under one manifest, one budget and one policy. A one-agent fleet is the simple case: an SEO agent on one site, or a nightly report. When the work is wider than one agent, the same fleet can run several, still under one cap. See fleet in the glossary.
Bring your own agent
- Claude Code
- Codex
- Hermes
- OpenCode
- Custom container
- Your own endpoint
Whatever the runtime, the same budgets, permissions, approvals and receipts apply. Model calls use your own key from any provider or OpenRouter.
01 · Budgets
A dollar budget that hard-stops spend
Set a monthly cap and a per-run cap in dollars. Every model and tool call spends from a small lease taken from the fleet's budget, so the worst overrun is one lease. At the cap, agents pause. They are never deleted, and they resume when you raise it.
- Monthly and per-run caps in dollars
- Warnings at 50, 75 and 90 percent
- Paused, never deleted, at the cap
02 · Schedules and triggers
Runs start on a schedule or a signal
Run hourly, nightly or on a cron expression, or start a run from Slack, a webhook, the API, or a chat in ChatGPT or Claude. Between runs an agent sleeps instead of holding a machine.
- Cron schedules
- Slack, webhook and API triggers
- Start runs from ChatGPT or Claude
03 · Sandboxes
A sandbox per agent
Each agent runs in its own sandbox. Model keys and connector credentials are attached outside it at the gateway, and agents get short-lived, narrow credentials, never your keys. Network egress leaves only through the gateway.
- One sandbox per agent
- Keys stay outside the sandbox
- Egress only through the gateway
04 · Approvals
Approvals for risky actions, not every action
Every tool call has an effect class: pure, reversible, compensatable or irreversible. Low-risk work runs within policy. Irreversible actions wait in escrow for a person, in Slack, ChatGPT, Claude or the console, and any approval can be saved as a rule.
- PureAgent, freely
- ReversibleAgent, within policy
- CompensatablePolicy, often a person
- IrreversibleAlways a person
05 · Audit
A tamper-evident audit trail with receipts
Every step is written to an append-only log where each event carries the hash of the one before it. Every run ends with a receipt: which agent, model and policy version, which checks passed, and the trace that led there.
- Hash-chained event log
- A receipt for every run
- Export for your own records
Keep reading
- How it worksHow Heighliner runs an agent fleet: describe it in a manifest, plan and quote every change, apply it, and govern risky actions with approvals.
- SecurityHow Heighliner contains AI agents: a sandbox per agent, keys outside the sandbox, egress through a gateway, approvals with escrow and a hash-chained audit trail.
- FleetA fleet is one or more AI agents running one harness under one manifest, one budget and one policy. A single scheduled agent is a fleet of one.
- Budget leaseA budget lease is a small slice of a fleet's dollar budget, $1 by default, that the gateway spends from locally, so caps are enforced on every call without slowing it.
Early access
Give your agents a budget, a sandbox and a supervisor.
Heighliner is in early access. Join the waitlist and we will reach out as places open.