Hard-stop budgets for AI agents: spend that stops at your cap
Autonomous agents can loop, retry, and burn tokens faster than you can notice. The fix isn't hoping they behave — it's a cap enforced outside them. In PAI the model gateway enforces per-Pie and per-company caps today; token salaries are designed to add a warning at 80% and an automatic pause at 100% (enableable), with every budget event signed into the audit chain.
Six controls, one hard limit
- Token salariesDesigned to give each Pie a spend allowance you set — nobody gets a blank cheque. Enableable.
- Hard stop at 100%Built to warn at 80% and auto-pause at 100%, on top of the live gateway cap that denies calls over budget.
- Cost on the chainEvery budget event is signed into a tamper-evident audit chain you verify offline.
- BYOK, at costBring your own keys; spend hits your provider account directly, with no markup.
- Per-Pie, per-companyBudgets nest: every Pie and company gets its own cap, one rollup for you.
- Approvals in chatTop up or raise a cap from the same thread you approve actions in.
Why AI agent costs run away
The same autonomy that makes an agent useful makes its spend unpredictable: a stuck task retries, a research loop fans out, a misfired webhook fires a thousand times. Real LLM cost control isn't a dashboard you check later — it's a limit the agent hits.
80% warn, 100% auto-pause
Underneath everything, the gateway cap is live: a call that would take a Pie or a company over budget is denied before it runs. Token salaries (enableable, switched on during setup) are designed to add two thresholds. At 80% of its salary a Pie sends a soft warning — top up, raise the cap, or let it ride. At 100% it is built to auto-pause and not spend again until you act. One is a notification; the other is a wall.
The MOAT: budgets you can prove
Here's what raw orchestration can't match: every budget event is signed into a tamper-evident audit chain you can verify offline. A budget the runtime merely reports is one you trust; a budget whose every breach is signed is one you can prove.
Keep reading
Budgets are one control. These cover the rest of the leash.
AI CEO Autopilot
Set a goal; an AI CEO staffs and runs it — and asks before it spends, every time.
Managed AI agents that don't get stuck
Stall detection and run contracts stop the loops that drive runaway spend in the first place.
Run multiple businesses with one AI workforce
Per-company budgets with a single rollup — isolation enforced by the database.
AI agent budgets — FAQ
Give it a cap enforced outside the agent. In PAI, per-Pie and per-company budgets are enforced at the model gateway — live today — so a call over the cap is denied before it runs. On top, token salaries are designed to warn you at 80% and auto-pause the Pie at 100% (enableable, switched on during setup). Because the cap sits at the gateway — not in the agent's good behaviour — a runaway loop is designed to stop at your cap instead of running on.
The soft warn (80%) is a heads-up: a message so you can top up, raise the salary, or let it ride. The hard stop (100%) is built to be structural: the Pie pauses and can't spend another token until you act. One is a notification; the other is a wall. Both come with the token-salary layer, which is enableable; the gateway cap underneath is live.
Yes. Each Pie has its own salary, and each company has its own cap with a single rollup for you. If you run three businesses, every business has an isolated budget and the family company can have the strictest of all. A cost-spike anomaly rule watches across the board for sudden jumps.
They make it honest. You bring your own AI provider keys, so spend goes to your provider account at cost — no token markup, no middleman. The budget is denominated in your real spend, not a marked-up credit. Local models (for example on a PAI Box) cost nothing per token, so work you send to them doesn't touch the cloud budget. Automatic privacy routing — sending prompts that carry personal data to the local model for you — is on the roadmap: planned, not live yet.
The hard stop is. Gateway caps (per Pie and per company) and the cost-spike anomaly rules are live and tested in our stack. The token-salary layer with the 80% soft-warn and 100% auto-pause rides the workforce layer and is enableable — switched on during setup. Either way, every budget incident is signed into the tamper-evident audit chain as an integrity event.
Govern spend structurally, not by hoping
See hard-stop budgets alongside every other capability — and the honest status of each — then choose your plan.