Priced per connected user. Not per tool call.
Arc0 charges for the users your agent works for, not for every search, retry and read along the way. Launch teams agree their price up front and keep it for 12 months.
Start free. Pay for active users in production.
Personal
Your own apps in every AI you use.
- ✓One Arc0 MCP URL for Claude, ChatGPT, Cursor
- ✓Your own connections, scoped per assistant
- ✓Audit log of every action, 7 days
- ✓Token export, any time
Build
Everything you need to ship your first agent.
- ✓Unlimited development-mode users
- ✓Up to 100 active connected users in production
- ✓Arc0 MCP, SDK, REST and proxy
- ✓Arc0 Connect with your logo and colours
- ✓Policies, approvals and audit log, 30 days
- ✓Token export, any time
Launch
RECOMMENDEDFor agents your customers depend on.
- ✓Everything in Build, no user cap
- ✓Your price locked for 12 months
- ✓White-label connect on your own domain
- ✓Your own OAuth apps, set up with you
- ✓Audit log, 90 days, with OpenTelemetry export
- ✓Triggers and webhooksSOON
Scale
For regulated and high-volume teams.
- ✓Everything in Launch
- ✓Custom retention and volume pricing
- ✓SSO and SCIM for your teamSOON
- ✓EU data residencySOON
- ✓VPC or self-hosted deploymentSOON
- ✓Uptime and support SLAsSOON
The same agent task, two ways to bill it.
| WHAT HAPPENS | PER-CALL PRICING | ARC0 |
|---|---|---|
| An agent task that takes 7 tool calls | 7 billable calls | 1 active user this month |
| A retry after the app times out | Billed again | Not billed |
| A call blocked by your policy | Billed on some platforms | Not billed |
| Testing in development | Counts toward the quota on some | Free |
| The same user, 200 tasks in a month | Every call of every task | Still 1 active user |
Why the unit matters for agents.
Per-call pricing made sense when integrations were called by code written once: a sync job, a webhook handler. Agents are different. They search before they act, read more than they need, retry on errors and fan out across apps. Newer patterns like code execution make one task into dozens of calls.
Pricing on tool calls turns every one of those steps into a cost you have to engineer around. Pricing on the user your agent works for tracks what you actually sell, stays predictable as your agent gets more capable, and never punishes a retry. If you want to see how it works end to end, start with how Arc0 works.
Pricing questions.
What is an active connected user?
An end user of yours with at least one action through Arc0 in the billing month. Users who connected apps but whose agent did nothing that month are not counted.
Why not charge per tool call?
Because agents make many calls per task: searching, reading, retrying, drafting. Per-call pricing punishes exactly the behaviour that makes agents useful, and code-execution modes multiply calls further. The user your agent works for is the unit that tracks the value you get.
What does fair use mean?
Rate limits per user and per workspace that normal agent traffic never reaches. If you have an unusual workload, such as bulk backfills, tell us and we will size it with you rather than bill you by surprise.
Are audit logs and white-label really on every plan?
The audit log and token export are on every plan, including free. Your logo and colours are on Build; your own domain and your own OAuth apps come with Launch. None of it is held back for an enterprise tier.
How is Launch priced?
We quote a price per active connected user from the volume you expect, before any of your users go live. That price is locked for 12 months.
Do my end users pay anything?
No. Your users connect their apps for free. Arc0 bills you, the company shipping the agent. Individuals using Arc0 for their own AI assistants use the free Personal plan.
Start free. Get a quote when you launch.
Tell us what your agent does and roughly how many users will connect apps. We will agree your price up front, and it stays locked for 12 months.
No card for Build · Failed calls free · Audit log on every plan