Priced per MCP call, not per seat.
A call is one request your agent makes to the endpoint — a tool search, a schema read, or a tool run. Free stops at its limit; every paid plan keeps working and bills the overage.
Trying Omniio, personal agents
- 25,000 MCP calls a month
- Calls stop at the limit — never a surprise invoice
- Every server in the library
- Search-first tool discovery
- 7 days of tool-call audit trail
- Signed webhooks and an API to read it back
Developers, power users
- 100,000 MCP calls a month
- Then €0.40 per 1,000 calls
- Every server in the library
- Search-first tool discovery
- 30 days of tool-call audit trail
- Signed webhooks and an API to read it back
Apps, small teams
- 500,000 MCP calls a month
- Then €0.25 per 1,000 calls
- Every server in the library
- Search-first tool discovery
- 3 months of tool-call audit trail
- Signed webhooks and an API to read it back
Production workloads
- 2,000,000 MCP calls a month
- Then €0.20 per 1,000 calls
- Every server in the library
- Search-first tool discovery
- 6 months of tool-call audit trail
- Signed webhooks and an API to read it back
- A team of up to 50 on one shared allowance
- Sign-in through your own Okta or Entra, with SCIM provisioning
Larger organizations
- Negotiated call volume
- Volume pricing
- Invoiced annually
- Every server in the library
- Search-first tool discovery
- 1 year of tool-call audit trail
- Signed webhooks and an API to read it back
- A team of any size on one shared allowance
- Sign-in through your own Okta or Entra, with SCIM provisioning
Tell it your volume. It does the sums.
Every figure below is computed from the same plan definitions the endpoint bills against, including the overage rate — so this is what the month would actually cost, not an illustration.
Inside the 100,000 calls this plan includes, so nothing is added.
Every plan, side by side.
| Plan | Monthly price | Included MCP calls | Overage | Burst limit | Audit retention | Team members | SSO & SCIM | Best for |
|---|---|---|---|---|---|---|---|---|
| Free | €0 | 25,000 | Hard limit | 60 / min | 7 days | — | — | Trying Omniio, personal agents |
| Pro | €19 | 100,000 | €0.40 / 1K | 120 / min | 30 days | — | — | Developers, power users |
| Scale | €49 | 500,000 | €0.25 / 1K | 300 / min | 3 months | — | — | Apps, small teams |
| Business | €149 | 2,000,000 | €0.20 / 1K | 600 / min | 6 months | Up to 50 | Included | Production workloads |
| Enterprise | Custom | Custom | Volume pricing | Custom | 1 year | Unlimited | Included | Larger organizations |
What counts, and what happens next.
What counts as a call
Each request to one of the three meta-tools — search_tools, describe_tool and run_tool — counts once. Connecting a client, listing your library, and browsing the app cost nothing. Your live counter is in the dashboard.
Reaching the limit
On Free, calls are refused once the 25,000 are spent and the counter resets on the 1st, UTC. On a paid plan nothing stops — calls past the included volume bill at that plan’s rate.
Before you pick a plan.
What counts as an MCP call?
One request to one of the three meta-tools: search_tools, describe_tool or run_tool. Connecting a client, authorising a server, browsing your library and reading the dashboard all cost nothing.
Do I need a card to start?
No. Free needs nothing but an email address and never asks for a card. A card is only needed when you choose a paid plan, and it is entered on Stripe’s own checkout page — Omniio never sees or stores your card details.
What happens when I use all 25,000 free calls?
Free is a hard limit: further calls are refused until the counter resets at the start of the next month, UTC. Your servers, credentials and library are untouched — only the endpoint stops answering. Paid plans never stop; they bill the excess at the plan’s rate.
Is there a limit on how fast I can call?
Yes — 60 / min on Free, rising with the plan and negotiated on Enterprise. It is a burst ceiling, not a second allowance: it sits far above what steady use of a plan’s monthly volume works out at, so what it catches is an agent stuck in a loop. A throttled call is refused with the seconds to wait and costs nothing against your monthly calls. The REST API is metered separately, per key rather than per account, so a script reading your audit trail can never throttle your agents.
Which servers do I get on each plan?
All of them. Every plan reaches all 145 servers in the library; the plan decides how many calls you make, never which tools you may reach.
Are my API tokens shared with anyone?
No. A credential you give Omniio is encrypted with AES-256-GCM before it is stored, decrypted only in memory to serve your own call, and never used for another account. Disconnecting a server deletes it. The privacy policy has the detail.
Can I see what my agent actually did?
Yes — every call is recorded with its tool, arguments, result and duration, and it is all in your activity log. Records are kept for 7 days on Free and up to 1 year on Enterprise, then deleted.
Can I get that record into my own systems?
Yes, on every plan, two ways. A signed webhook posts each call to a URL you own as it finishes, with an HMAC your receiver can verify. Or poll the read API with an API key on your own schedule. Omniio also emits OpenTelemetry spans and joins the trace your client is already running if it passes one.
Can my colleagues share one plan?
Yes, on Business and Enterprise. Create a team in your team settings and invite people by email; everyone draws on the one monthly allowance and there is one invoice. A team shares the plan and nothing else — each person keeps their own servers, their own credentials and their own MCP endpoint, so joining a team never hands anyone your API tokens.
Can I change or cancel a plan later?
At any time, from the billing portal in your settings. Upgrades apply straight away and Stripe prorates the difference. A cancellation keeps the plan running until the end of the period you have paid for, then drops the account back to Free and its hard limit.
Start on Free. Move when it pays.
25,000 calls a month, every server in the library, and the same endpoint you’d use on any other plan.