ooligo

Arcade

agent-integration-platform mcp-gateway · agent-authentication · tool-calling · agent-governance
AI-NATIVE MCP API FREEMIUM
RevOpsLegal OpsRecruiting & TACustomer Success
7.8 /10

What it is

Arcade calls itself “the actions runtime” between your agents and the systems they reach: the layer that decides whether a given agent may take a given action, on whose authority, and records that it did. Where Composio competes on the lowest price per call, Arcade competes on a different question: which human authorized this write, and can you prove it six months later. Authorization is delegated, so the agent acts as its user. When a tool call needs an OAuth scope the calling user has not granted, Arcade pauses the run, sends that user through the provider’s consent flow, and resumes. The token in play belongs to the person making the request, not to one shared platform key. It works through your existing OAuth and identity-provider flows (Okta, Entra) instead of replacing them.

The company raised a $60M Series A in June 2026 led by SYN Ventures, with Morgan Stanley and Wipro investing strategically, for $72M in total funding including a $12M seed. It wrote the MCP tool-authorization proposal behind the protocol’s URL elicitation flow, developed with Anthropic and accepted into the spec in November 2025. On 5 August 2026 it acquired Smithery, the public MCP registry and hosting platform. Production customers named on its homepage include LangChain, Snyk, Relevance AI, Sybill, Duracell and Madrigal Pharmaceuticals.

Why it shows up in agent stacks

One platform OAuth client is one blast radius. Composio disclosed on 21 May 2026 that attackers took roughly 5,241 API keys and 5,001 GitHub OAuth tokens from exactly that kind of central store. With Arcade’s per-user model, the same incident reaches only the grants of the individual accounts that were hit. It also leaves an action log naming the human who authorized each write. A compliance reviewer asks for that log, and a shared service account cannot produce it.

Read the catalogue numbers carefully. Arcade advertises “8,000+ permission-aware tools,” but its public catalogue at arcade.dev/tools lists about 85 named servers in nine categories, from Salesforce, HubSpot and Attio to Zendesk, ServiceNow, the Microsoft 365 suite and Workday. A tool is one function; a server is what you actually integrate against. Some cards carry a Pro badge and some a BYOC (bring your own credentials) badge, and a separate tier of partner-built servers sits beside Arcade’s own.

Pricing reality

Arcade dropped auth-event billing. The live pricing page meters one thing, tool calls, and says plainly “not per seat”:

  • Free: $0, no credit card, 2,000 tool calls a month on Arcade Cloud, community support through Discord and GitHub.
  • Team: $25/mo platform fee plus $0.01 per tool call, Arcade Cloud only, next-business-day email support. The page lists no included call allowance on Team.
  • Enterprise: custom, sold as annual bundles of tool calls with discounted overage as you scale. This is the only tier with SSO, RBAC, audit logs, private registry access, deployment in your VPC or fully air-gapped, 24/7 SLA-backed support, and a dedicated forward-deployed engineer.

The real band is now simple arithmetic. An internal agent serving 300 reps at 25,000 tool calls a month costs $275 ($25 platform plus $250 in calls), whether those calls come from 30 users or 3,000. At 500,000 calls a month, Team list price is $5,025, which is the volume where you stop paying list and negotiate a bundle. The old penalty for large rollouts is gone. Under the previous meter, a 5,000-seat deployment across four providers billed 20,000 auth events before the agent did any work. Today headcount costs nothing; only actions do.

Per call, $0.01 against Composio’s $0.0003 is still a 33x gap. That premium buys per-user tokens and a per-action audit trail. If nobody will ever ask which human authorized a write, you are paying more than an order of magnitude extra for a property you do not use.

Best for

The platform or security engineer shipping an agent that writes to a system of record on behalf of many named end users, where the audit requirement is “prove which human authorized this action.” The scoped case where it wins outright: a RevOps or Customer Success team putting an agent in front of several hundred reps that updates Salesforce or HubSpot and posts to Slack as each rep. Every write keeps a per-user trail that survives a compliance review. With seat-driven auth billing gone, the bill tracks the actions the agent takes, not how many reps it serves.

Not for

A single-tenant internal agent running as one service account at high call volume. If the same three tokens serve every request and no reviewer will ever ask whose authority a write carried, the 33x per-call premium buys nothing. It is also the wrong layer for a non-engineer: this is an SDK, a CLI and an API, not a canvas. And every governance control a legal-ops or recruiting reviewer will ask about, from audit logs and SSO to VPC deployment, sits behind Enterprise. Team is not a compliant production tier for regulated data.

Versus the alternatives

  • Composio: the cheapest authenticated tool call in the segment at $0.0003, with 100,000 free calls a month. Pick it when call volume dominates the bill and a single platform key is custody your security review will accept. Its May 2026 breach is the argument against that acceptance.
  • Zapier MCP: the largest install base in the category at 9,000+ apps, included on every Zapier plan. Pick it when an ops generalist does the wiring, monthly volume stays under about 1,000 tool calls, and Zapier is already on the invoice. Task-based billing works out near $0.069 per call, so the advantage flips the moment volume grows.
  • Runlayer: the fastest-growing entrant in agent governance, with an $11M seed in November 2025 and a $30M Series A in June 2026 led by Felicis. It sells quote-only as a security platform. Pick it when the problem is discovering shadow MCP servers and inspecting traffic from servers you did not build. Arcade wins on identity, Runlayer on inbound threat inspection, and teams with both problems buy both.
  • Merge Agent Handler: managed auth plus DLP on every call, from $1,000/mo. Pick it when data-loss prevention on employee agent traffic is the requirement and your HRIS, ATS and CRM already connect through Merge’s single API.

If none of these fit, the honest fallback below about five integrations and one service account is writing direct API clients and skipping the layer.

Watch-outs

  • The homepage says “SOC 2 compliant” without naming the type, and there is no ISO 27001 or HIPAA position anywhere on the public site. Guard: ask in writing for the SOC 2 report, its type, the audit period and the latest pen-test summary before the security review starts. Treat the marketing line as unverified until the document is in hand. A legal-ops or recruiting deployment touching contracts or candidate PII will not clear review on a homepage claim.
  • Pro-badged tools may not bill at $0.01. Arcade’s 2025 pricing charged $0.50 per Pro execution, or $0.01 if you brought your own credentials. The current pricing page names no Pro rate, but the badge is still on the catalogue. Guard: before you build on a Pro-badged server such as the search or scraping tools, get the per-call rate confirmed in writing and run a metered test week on Team to see what actually lands on the invoice.
  • The Free tier is a demo, not a pilot. 2,000 tool calls is one afternoon of a chatty agent, and the ceiling arrives without warning mid-evaluation. Guard: run any pilot that ends in a go/no-go decision on Team with an explicit budget cap, so the evaluation ends on evidence rather than on a quota.
  • A large catalogue is not a vetted one. Arcade’s own index of 43,400+ public MCP servers and 219,069 tools found that 0.5% earned an A grade, and its own catalogue mixes Arcade-built, partner-built and BYOC servers. Guard: test the three tool calls your workflow depends on against live data during the trial, and confirm who maintains each server you rely on, rather than trusting the catalogue entry.