Arcade.dev

Arcade.dev is a Agent tool authorization product worth evaluating when you have a concrete workflow, not just a vague mandate to 'add AI.' The current vendor research was based on curl fetches of the homepage and pricing page, plus available search-result text. The useful evidence was specific: pricing page listed Hobby Free, Growth $25/month plus per-use charges, Enterprise custom, MCP compatibility, user challenges, standard/pro tool executions, hosted MCP servers, RBAC, SSO, and SAML. That makes the product easiest to judge around five practical capabilities: MCP runtime designed for secure production AI agent deployments; Connects to identity providers, enforces agent authorization, and enables actions in Google, Slack, Salesforce, and other tools; OAuth handling, permission challenges, hosted MCP servers, audit logs, RBAC, SSO, and SAML on higher tiers; Usage-based pricing based on user challenges, standard tool executions, pro executions, and hosted MCP server hours; Developer resources include docs, sample apps, tool patterns, LangSmith, and Arcade Tools. Builders should test those capabilities with production-shaped inputs, because AI demos often hide the real costs: setup time, review time, integration friction, and failure cases.

Pricing matters here. The researched pricing snapshot is: Hobby Free — 100 user challenges, 1,000 standard tool executions, 50 pro tool executions, MCP compatibility, 1 Arcade-hosted MCP server, and GitHub community support.; Growth $25/month + usage — 600 user challenges then $0.05 each, 2,000 standard tool executions then $0.01 each, 100 pro executions then $0.50 each, hosted MCP servers at $0.05/hour, and email SLA support.; Enterprise Custom — Dedicated infrastructure, custom pricing, dedicated account representative, custom SLAs, tenant isolation, audit logs, RBAC, SSO, and SAML.. Do not treat that as a procurement quote; treat it as enough context to decide whether this belongs in a free experiment, a small team pilot, or enterprise buying. The vendor pages exposed enough readable evidence to summarize pricing and core capabilities, but buyers should still confirm plan limits before signing. If usage is metered, model the cost around your real volume: minutes of video, tool executions, memories, users, or engineering tasks per month.

The strongest reasons to shortlist Arcade.dev are: Pricing clearly exposes the usage meters that matter for agent actions; Strong security posture for permissioned tool use and OAuth-heavy workflows; Good fit when agents need to take real actions rather than only answer questions. Arcade.dev is different because it treats authorization and secure action execution as the core product, not an afterthought on top of connectors. Those strengths are most valuable when the work repeats every week and when the team can define an acceptable output. The best pilot is narrow: pick one workflow, run 20 to 50 representative tasks, track the percentage accepted without edits, and record how often a human has to step in.

The trade-offs are just as important: Usage-based pricing requires careful modeling for high-volume agents; Adds infrastructure complexity if a team only needs a simple internal prototype; Production value depends on correctly designing scopes, approvals, and audit workflows. These are not deal-breakers, but they are the areas that decide whether the tool saves time or creates a new review queue. For customer-facing, regulated, or revenue-critical workflows, keep human approval in the loop until error rates are known. Also verify admin controls, data retention, SSO or OAuth behavior, export rights, and support response times.

Best-fit use cases include Permissioned agents: Let assistants act in SaaS tools while challenging users for sensitive permissions.; Enterprise workflow agents: Add RBAC, audit logs, tenant isolation, SSO, and SAML around agent actions.; MCP deployment layer: Host and operate MCP-compatible tool access for production applications.. My practical recommendation: run a 7 to 14 day pilot against one existing process, compare Arcade.dev with at least two adjacent tools, and buy only if it produces a measurable win: lower cost per output, faster cycle time, better consistency, or a workflow the team could not realistically operate before.