Skip to content

Pricing

$0.20 per 1,000 checks

One check is one tool call judged on all nine questions. Every workspace starts with 1,000 free requests; then buy packs as you go or pick a monthly plan with bonus checks. Failed calls are never billed.

Pay as you go

$0.20/ 1,000 checks

  • 1,000 free requests
  • Nine scores per check, one price
  • Batch up to 64 checks per request
  • Failed calls are free

Packs and plans

Buy as you go, or subscribe for bonus checks

Packs are the list rate and never commit you to anything. Plans renew monthly and include +5%, +10%, +15% extra checks.

Packs

$5

25,000 checks

$0.20 per 1,000 checks

$25

125,000 checks

$0.20 per 1,000 checks

$100

500,000 checks

$0.20 per 1,000 checks

Monthly plans

Starter

+5% checks

$19/ month

99,750 checks a month

$0.19 per 1,000 checks (vs $0.20 list)

For a team guarding its first agents in production.

Start with Starter

Pro

+10% checks

$99/ month

544,500 checks a month

$0.18 per 1,000 checks (vs $0.20 list)

For products that run agents for their customers.

Start with Pro

Scale

+15% checks

$399/ month

2,294,250 checks a month

$0.17 per 1,000 checks (vs $0.20 list)

High-volume agent fleets, priority capacity.

Start with Scale

Calculator

What will my agents cost to guard?

Estimate your checks per month (roughly: tool calls per agent run × runs per month) and compare pay-as-you-go with each plan.

1k50M

1 check = one tool call judged on all nine questions = 1 credit, at $0.20 per 1,000 checks. A batch of N checks costs N credits. Failed calls are free, and every workspace starts with 1,000 free requests.

For reference, illustrative

An LLM judge on every call (a general model asked “is this safe?”) costs roughly $2–$5 per 1,000 calls and adds ~1 s per call, depending on the model and prompt size. At this volume that is about $200–$500 a month. A rough market range, not a quote: your numbers will differ.

Cheapest option · Starter

$19.05/ month

$19 plan + 250 extra checks at the list rate

  • Starter$19.05
  • Pay as you go$20
  • Pro$99
  • Scale$399

For reference

A small guard model versus an LLM judge

An illustrative comparison, not a benchmark: LLM prices and latencies vary widely by model, provider and prompt size.

MeasureMCP GuardLLM judge (illustrative)
Cost per 1,000 calls$0.20~$2–$5
Added latency per call~15 ms model time + network~1 s
OutputVerdict + nine calibrated scoresFree text you parse into a decision
StrengthCheap and fast enough for every callDeeper reasoning on unusual cases

They combine well: run the guard on every call and send only the ask verdicts to an LLM judge or a person.

FAQ

Pricing questions, answered

What am I billed for?
Checks. 1 credit = 1 check = one tool call judged on all nine questions in one forward pass, at $0.20 per 1,000 checks. A batch request with N checks costs N credits. Only successful checks are billed.
What are the free requests?
Every workspace gets 1,000 free requests, once: a lifetime total, never reset monthly. A free request costs nothing whatever it contains, and a batch request counts once. After that you buy a pack or subscribe to a plan.
Are failed calls billed?
No. Validation errors (400), out-of-credit responses (402), payload-too-large (413), rate limits (429) and unavailable capacity (503) cost nothing and do not use a free request.
Is there a surcharge for Swiss hosting?
No. Checks run on our GPUs in Switzerland at the same price, and Swiss-only processing (no EU failover) costs nothing extra.
Which option should I pick?
Start with the free requests. For small or irregular volume, buy a pack (from $5); packs are the list rate. Plans include up to 15% more checks for the same money, so they pay off once you use most of the monthly allowance. The calculator shows the cheapest option for your volume.
How does this compare with using an LLM as a judge?
As a rough, illustrative reference: asking a general LLM “is this tool call safe?” on every call costs roughly $2–$5 per 1,000 calls and adds ~1 s per call, depending on the model and how much context you send. MCP Guard is $0.20 per 1,000 checks and about 15 ms of GPU time per check. An LLM judge can reason about unusual cases in more depth; many teams use a small guard on every call and escalate the uncertain ones.
Is there annual billing, and how do plan changes work?
Yes: every plan is also sold yearly, 12 months for the price of 11 (Starter $209/yr, Pro $1089/yr, Scale $4389/yr). The monthly checks are added every month of the paid year, not all at once. An upgrade applies immediately: you pay the prorated difference for the rest of the billing period and get the prorated extra checks. A downgrade applies at the end of the billing period, with no charge and no refund.
Do you offer volume pricing or dedicated capacity?
Email support@mcp-guard.ai with your expected volume and we will talk it through.

Which errors are free is listed in the docs.

Try it on your own agents

$0.20 per 1,000 checks · 1,000 free requests