AI spend control gateway

Your team ships AI. You control the bill.

One gateway for OpenAI and Anthropic. Hand engineering, agencies, and ops virtual keys with hard budgets, per-client cost reports, and one-click revocation.

Start free Read the docs

No credit card · First governed request in under 5 minutes

The whole migration

client = OpenAI(
  api_key="kc-…",
  base_url="https://api.keyc.ai/v1"
)

<5 min

signup → first governed request

<30 ms

p95 added latency

prompts stored — ever

1 click

revoke any key globally

Every request, governed in flight

Apps use virtual keys. KeyC checks budget and rules, swaps the encrypted provider key, forwards the call, and meters cost — in milliseconds.

Built for every team

Marketing that speaks your team's language

Same gateway — different wins. Pick the story that matches who signs the check, who writes the code, and who answers the pager.

For Engineering

Developers & platform teams

Ship fast without handing out the master key

Mint a virtual key per service, developer, or environment. Model allowlists, RPM limits, and expiry dates — without a secrets rotation fire drill.

Drop-in OpenAI & Anthropic SDK — change base_url + api_key only
Streaming, retries, and tool calls work unchanged
Revoke one key in one second across every gateway replica
Never commit sk-… again — leak a kc- key, kill just that key

Start building

Engineering team using virtual API keys through KeyC

For Finance & Ops

CFO, FP&A, and ops leads

Hard caps beat surprise invoices

Set monthly budgets per key before the spend happens. At 100%, requests stop with a clear 402 — not a five-figure bill at month-end.

Budget enforcement on every request, in milliseconds
Live spend vs cap on one screen
Automatic reset on the 1st (UTC) — predictable cycles
You still pay providers directly — KeyC never marks up tokens

Control spend

Finance team monitoring AI budgets on KeyC

For Agencies & consultancies

Client delivery & account teams

Every client gets their own key — and their own receipt

Ten clients on one OpenAI account used to mean one mystery invoice. Tag keys by client, cap each budget, export line-item reports accounting can use as-is.

One virtual key per client or project
Monthly cost breakdown by tag — export-ready
Project ends? Set expiry — access dies automatically
Bill pass-through AI costs with proof, not spreadsheets

Bill with confidence

Agency team billing AI costs per client with KeyC

For Security & compliance

CISO, security engineers, GRC

Govern access without reading prompts

Real provider keys live encrypted in the vault. Virtual keys are hashed — shown once, never stored. We meter metadata only: tokens, model, latency, cost.

AES-256-GCM for provider keys; master key stays in your env
Full audit trail of key usage — no prompt content logged
Instant revocation propagates via Redis pub/sub
Built for teams that need control, not another data processor

See privacy model

Security-focused vault and compliance with KeyC

For DevOps & platform

SRE, platform, and on-call

Alerts at 80%. Blocks at 100%.

Budget thresholds hit Slack or email once — not as a pager storm. Gateway replicas stay hot; Postgres never sits on the request path.

Slack + email alerts on configurable thresholds
Sub-30ms p95 added latency — proxy built for production
Redundant gateway replicas behind your edge
Status and enforcement metrics you can trust on-call

Run in production

Operations team receiving KeyC budget alerts

Six things your provider dashboard won't do

Governance built for production AI — not a billing afterthought.

Virtual keys

One vault. Infinite governed copies.

Encrypted provider keys stay put. Every developer, client, or cron job gets a kc- key with its own rules.

Hard budgets

Stop at the limit — not at the invoice.

Monthly caps enforced per request. HTTP 402 when spent — SDK-friendly, graceful, predictable.

Per-client billing

Every dollar has a name on it.

Tag keys, watch live burn, export monthly breakdowns your finance team can reconcile.

Instant revoke

Offboard in one click.

Revoked keys die everywhere in about a second. No rotation ceremony, no redeploy.

Smart alerts

Slack at 80%. Block at 100%.

Threshold notifications that respect on-call — actionable, not noisy.

Private by design

We never store prompts.

Metadata-only metering. Your completions pass through memory and leave no trace in our logs.

Simple pricing. Your tokens stay yours.

Pay OpenAI and Anthropic directly. KeyC is a flat subscription for control — never a markup on usage.

Free

Side projects & evaluation

$0/mo

2 virtual keys
10K requests / month
Budgets & model allowlists
Community support

Start free

Pro

Teams shipping AI to production

$29/mo

Unlimited virtual keys
100K requests / month
Slack + email alerts
Rate limits & key expiry
Priority support

Start Pro

Team

Agencies billing clients

$99/mo

1M requests / month
Client tags & cost reports
Monthly per-client export
Team member accounts
Shared Slack channel

Start Team

Enterprise volume, SLA, or on-prem? founders@keyc.ai

Questions, answered

How long does setup really take?+

Under five minutes. Vault your provider key, mint a virtual key, change api_key and base_url in your SDK. Streaming included.

Do you store my prompts?+

Never. We log metadata only: token counts, model, latency, cost, and status. Content passes through in memory only.

What happens when a budget runs out?+

HTTP 402 in standard OpenAI error shape. Budgets reset on the 1st (UTC). Raise a cap anytime — it applies instantly.

How is this different from LiteLLM?+

LiteLLM is powerful self-hosted infra. KeyC is the managed spend-control layer with per-client billing — live in minutes, not sprints.

Do you charge for my token usage?+

No markup. You bring provider keys and pay OpenAI/Anthropic directly. KeyC is a flat subscription for governance.

What if KeyC goes down?+

Redundant replicas, cache-first auth, and deliberate fail-open on brief internal issues — your app stays up while enforcement catches up.

Two lines of code between you and a bill you control

Engineering gets speed. Finance gets caps. Agencies get receipts. Security gets vaults. Ops gets alerts that matter.

Get started free