Arcade pricing: how do I estimate monthly cost from tool executions, user challenges, and workers?

Most teams estimating Arcade pricing care about one thing: “What will this actually cost me when my agents are live?” The answer comes down to three main usage drivers: user challenges (auth), tool executions (standard vs pro), and workers (Arcade-hosted vs self-hosted). Once you understand those levers, you can forecast monthly cost with pretty high confidence.

Quick Answer: Estimate your Arcade bill by adding: (1) plan base price, plus (2) overages on user challenges, standard tool executions, and pro tool executions beyond the included amounts, plus (3) any Arcade-hosted worker hours at $0.05 per server-hour. Self-hosted workers don’t add usage cost.

---

Frequently Asked Questions

How does Arcade pricing work at a high level?

Short Answer: Arcade uses transparent, usage-based pricing: you pick a plan (Hobby, Growth, or Enterprise), then pay only for usage beyond what’s included—user challenges, tool executions, and hosted worker hours. No seat licenses, no hidden fees.

Expanded Explanation:
Arcade’s pricing is designed around how real agents run in production. Instead of charging per user or per “seat,” you pay for three things your agents actually consume:

• Authentication events (“user challenges”): when an agent needs a user to grant or refresh access to a system via OAuth/IDP.
• Tool executions: when the MCP runtime actually calls a tool on your behalf (e.g., Google.SendEmail, Slack.PostMessage).
• Hosted compute (“workers”): when you run MCP servers or workers on Arcade-hosted infrastructure.

Each plan includes a bundle of usage to get you started:

• Hobby (Free)

  • 100 user challenges
  • 1,000 standard tool executions
  • 50 pro tool executions
  • 1 Arcade-hosted worker (included)
  • 5 self-hosted workers
  • Community support via GitHub

• Growth ($25/month + usage)

  • 600 user challenges included, then $0.05 each
  • 2,000 standard tool executions included, then $0.01 each
  • 100 pro tool executions included, then $0.50 each
  • Unlimited Arcade-hosted workers at $0.05 per server-hour
  • Unlimited self-hosted workers
  • Email support with SLA

• Enterprise (Custom)

  • Dedicated infrastructure, tenant isolation, audit logs, RBAC, SSO/SAML
  • Custom usage and volume pricing
  • Cloud, VPC, on-prem, or air-gapped deployments

Key Takeaways:

• Your bill is driven by usage, not user seats.
• Once you know your expected auth prompts, tool calls, and hosted worker hours, you can model cost very easily.

---

How do I estimate cost from user challenges, tool executions, and workers?

Short Answer: Count how many auth prompts (user challenges), standard tool calls, pro tool calls, and hosted worker hours you expect per month, subtract what’s included on your plan, and multiply the overage by the per-unit prices.

Expanded Explanation:
Think of Arcade pricing as a simple formula:

Monthly Cost = Plan Base + User Challenge Overage + Standard Tool Overage + Pro Tool Overage + Hosted Worker Hours

On Hobby, you only pay once you exceed the free included usage. On Growth, you pay $25/month plus overages. Enterprise is custom but follows the same usage patterns with volume pricing.

To estimate your cost, you’ll:

1. Pick a plan (most teams starting production use Growth).
2. Forecast how often your agents will need to challenge a user for auth.
3. Estimate how many tool executions they’ll perform, broken into standard vs pro.
4. Decide how many workers you’ll run on Arcade-hosted infrastructure, and for how many hours.

Steps:

1. List your agent scenarios (e.g., “daily email digests,” “Slack standup bot,” “GitHub triage agent”) and estimate how many times each runs per month.
2. For each scenario, count tool calls per run (e.g., list emails, summarize, send Slack message) and multiply by runs/month to get total tool executions.
3. Add in user challenges you expect (how often users will connect or re-auth to Gmail, Slack, GitHub, etc.).

Once you have those numbers, plug them into the plan’s included usage and per-unit rates.

---

What’s the difference between standard tool executions and pro tool executions?

Short Answer: Standard tool executions cover the majority of everyday actions (e.g., listing emails, sending Slack messages), while pro tool executions are advanced tools with higher infrastructure costs, priced separately at a higher rate.

Expanded Explanation:
Arcade splits tools into two buckets to keep core usage inexpensive and predictable:

• Standard tools – agent-optimized tools with typical infrastructure cost. Think:

  • Gmail.ListEmails
  • Google.Calendar.ListEvents
  • Slack.PostMessage
  • GitHub.ListIssues
  • Most routine “read” or simple “write” actions your agent uses frequently.

• Pro tools – advanced tools that are heavier on infrastructure (compute, storage, or specialized services). They’re built for more complex workflows and are priced higher:

  • Complex cross-system orchestrations
  • High-volume or long-running operations
  • Scheduled future executions (which add a small scheduling fee)

From the pricing context:

• Hobby: 50 pro tool executions included.
• Growth: 100 pro tool executions included.
• After that: $0.50 per pro execution (plus $0.005 if scheduled for later).

Comparison Snapshot:

• Standard Tools:
  • Included amounts: 1,000 (Hobby), 2,000 (Growth)
  • Overage: $0.01 per execution
  • Best for: high-frequency, everyday agent actions.
• Pro Tools:
  • Included amounts: 50 (Hobby), 100 (Growth)
  • Overage: $0.50 per execution (+$0.005 if scheduled later)
  • Best for: heavy, advanced, or scheduled operations where extra cost is tied to extra infrastructure.

---

How do hosted and self-hosted workers affect pricing?

Short Answer: Self-hosted workers don’t add usage cost; Arcade-hosted workers are billed at $0.05 per server-hour beyond the included number.

Expanded Explanation:
Workers are where your MCP servers actually run. You have two options:

• Arcade-hosted workers: zero setup, managed by Arcade. Ideal for fast starts and cloud-native deployments.
• Self-hosted workers: run the MCP runtime in your own infrastructure (Kubernetes, VMs, on-prem, or air-gapped), while Arcade still handles auth and tool coordination.

From the pricing details:

• Hobby

  • 1 Arcade-hosted worker included (24/7)
  • 5 self-hosted workers included

• Growth

  • Unlimited Arcade-hosted workers at $0.05 per server-hour
  • Unlimited self-hosted workers

• Enterprise

  • Unlimited workers with dedicated infrastructure and volume pricing

To estimate worker cost, you only need to think about Arcade-hosted workers:

Worker Cost = Number of Arcade-hosted workers × hours per month × $0.05

Example:

• You run 2 Arcade-hosted workers, 24/7:
  • 2 workers × 24 hours/day × 30 days = 1,440 worker-hours
  • 1,440 × $0.05 = $72/month for workers

Self-hosted workers are usage-free; your cost there is just your own infrastructure.

What You Need:

• A decision on whether you want zero-setup Arcade-hosted workers, or to run workers inside your own infra.
• A rough idea of how many workers you’ll keep online and whether they’re 24/7 or bursty.

---

How do I model a realistic monthly bill for my agents?

Short Answer: Start from your expected workflows, translate them into counts of auth challenges, standard tool calls, pro tool calls, and hosted worker hours, then apply the plan’s included usage and per-unit pricing to get a conservative estimate.

Expanded Explanation:
When I’ve sat in budget meetings as a CTO, the only estimates that survive are the ones tied directly to real workloads. For Arcade, that’s not “number of users,” it’s “what does the agent actually do?”

Here’s a concrete example on the Growth plan ($25/month):

Scenario:
You have one Gmail + Slack agent that:

• Sends a daily email summary to 100 users.
• Posts a daily Slack standup message to 5 channels.
• Runs 24/7 on Arcade-hosted workers.

Step 1: Estimate tool executions

• Email summary per user per day

  • Gmail.ListEmails (standard)
  • Google.SendEmail (standard)
  • 2 standard tool executions per user per day.
  • 100 users × 2 calls/day × 30 days = 6,000 standard executions

• Slack standup

  • Slack.PostMessage (standard)
  • 1 call/channel/day
  • 5 channels × 1 call/day × 30 days = 150 standard executions

Total standard tool executions: 6,150/month

Assume 0 pro executions in this example.

Step 2: Estimate user challenges

Let’s say each of the 100 users connects Gmail once in the first month:

• 100 OAuth flows = 100 user challenges

On Growth, you get 600 user challenges included, so you’re within the included amount.

Step 3: Estimate worker hours

You run 2 Arcade-hosted workers 24/7:

• 2 workers × 24 × 30 = 1,440 worker-hours
• 1,440 × $0.05 = $72

Step 4: Apply plan inclusions and overages

Growth plan gives:

• 2,000 standard executions included, then $0.01 each
• 600 user challenges included, then $0.05 each
• 100 pro executions included, then $0.50 each
• Unlimited Arcade-hosted workers at $0.05/server-hour

Now calculate:

1. Base plan: $25
2. Standard tool overage:
   • Used: 6,150
   • Included: 2,000
   • Overage: 4,150 × $0.01 = $41.50
3. User challenges:
   • Used: 100
   • Included: 600
   • Overage: 0 → $0
4. Pro tools:
   • Used: 0
   • Included: 100
   • Overage: 0 → $0
5. Hosted workers:
   • 1,440 hours × $0.05 = $72

Estimated Monthly Cost = $25 + $41.50 + $72 = $138.50

You’ve got a production Gmail + Slack agent running non-stop for ~100 users at a predictable sub-$150/month cost.

Why It Matters:

• You can scale agents without guessing; every extra tool call and worker hour has a clear marginal cost.
• Budget conversations shift from “What if it spikes?” to “If we double agent usage, cost doubles in a predictable way.”

---

Quick Recap

Arcade pricing is built around what actually drives cost for production agents: user challenges into services, standard vs pro tool executions, and Arcade-hosted worker hours. Plans (Hobby, Growth, Enterprise) give you generous included usage, then simple per-unit pricing once you cross those thresholds. To estimate your monthly cost, translate your workflows into counts of auth prompts, tool calls, and worker-hours, subtract the included amounts on your plan, and multiply the overage by the published rates.

Next Step

Get Started