Solutions
One gateway, four ways to protect margin.
The same request path serves founders worried about power users, platform teams running agents, agencies billing clients and finance teams closing the month. Pick the lens that matches your problem.
| Tenant | Plan | Plan price | LLM cost | Policy |
|---|---|---|---|---|
| acct_7Q2 northwind-legal | team | $99 | $412.60 | hard cap at $35 |
| acct_1C9 brightpath-hr | team | $99 | $31.20 | within allowance |
| acct_4K1 atlas-freight | pro | $249 | $198.40 | soft cap 80% |
| acct_9D3 lumen-dental | starter | $49 | $6.80 | within allowance |
By company and team
Start from the problem you have today.
- 01B2B AI SaaSFlat-rate plans with unmetered power users. Per-tenant caps and overages keep every plan above its margin floor.Quotas, billing sync
- 02Agentic platformsMulti-step agents that recurse, re-plan and thrash tools. Budget envelopes per run and four loop heuristics halt runaways in under 50 ms.Circuit breakers, envelopes
- 03AI development agenciesMany clients, shared provider keys, hand-built invoices. One workspace per client, budget caps, and markup billed automatically.Isolation, markup billing
- 04FinOps and financeOne provider invoice, no idea which customer it belongs to. Gross margin per tenant, per tier and per feature, every hour.Metering, margin reports
Who PALM SOUQ is for
Four customer shapes, four failure modes.
If none of these fit, for example a single-user app or fully self-hosted models with no token bill, PALM SOUQ is probably more than you need.
| Segment | Typical product | What breaks | PALM SOUQ lever | |
|---|---|---|---|---|
| GenAI and agentic SaaS, Seed to Series B | Chat, agents, code generation or document processing on flat or tiered plans | A few heavy tenants consume most of the token budget and turn their plans negative | Per-tenant caps, fallbacks, overages | Read more |
| Enterprise software adding AI assistants | Established product with a new AI add-on tier | Finance cannot see AI COGS per account, and liability for API spend is open-ended | Margin per account and feature, billing sync | Read more |
| AI development agencies and SIs | Bespoke agents built and run for several enterprise clients | Client costs are mixed on shared keys and re-billed by spreadsheet | Client workspaces, budgets, markup | Read more |
| Autonomous browser and coding agent platforms | Long-running agents with tools, sub-agents and retries | Recursion and retry loops can spend thousands before anyone looks | Loop heuristics, run envelopes, depth caps | Read more |
Who signs off
Every buyer gets a straight answer.
VP Engineering or CTO
Under 5 ms added latency, 99.999% availability, a one-line drop-in.
Edge gateway on 280+ PoPs, fail-open SDKs, base URL change.
Head of Finance or FinOps
Gross margin per customer, accurate token prices, billing sync.
Hourly margin by tenant and tier, cache-aware pricing, Stripe, Metronome, Orb and Lago.
Staff AI or backend engineer
Debugging loops, granular limits, multi-provider SDKs.
Breaker traces with the exact trigger, Python and TypeScript SDKs, OpenTelemetry export.
Chief Product Officer
Fair usage, graceful degradation, no surprise lockouts.
Soft caps, model fallbacks and upgrade prompts instead of hard errors.
Don't wait for a surprise $10,000 OpenAI invoice.
Route your agent traffic through PALM SOUQ in 2 minutes. Change one base URL, pass a tenant header, and every token is attributed and governed.
14-day free production trial · 100k requests included · no card required