A

Arcade.dev

by Arcade.dev

Agent DevelopmentGovernance & SecurityDeveloper ToolsAI Agents & Orchestration

The actions runtime that lets AI agents act as their users, with authorization, execution and audit built in.

Freemium · Usage-based · Contact for pricing·Added Jul 11, 2026·Updated Oct 1, 2026
Share:
THE DAILY BRIEF
Arcade.dev

by Arcade.dev

Agent DevelopmentGovernance & SecurityDeveloper ToolsAI Agents & Orchestration

The actions runtime that lets AI agents act as their users, with authorization, execution and audit built in.

Freemium · Usage-based · Contact for pricing

Arcade.dev is an authorization and tool-execution runtime for AI agents. It sits between agents and business systems such as Gmail, Slack, GitHub and Salesforce. It handles per-user OAuth, applies policy outside the model, and runs actions from a catalog of thousands of MCP tools. That lets engineering teams take agents from demo to production without building their own credential and permission plumbing.

At a Glance

Category
Agent Development
Pricing
Freemium, Usage-based, Contact for pricing
Target Market
CTOs, CIOs, Enterprise Developers, Security Teams, AI Platform Engineers
Deployment
Cloud-first, Self-hosted, Hybrid, API-based
Founded
2024
Headquarters
San Francisco, USA
Team Size
11-50

Key Features

  • ✓Delegated user authorization
  • ✓Permission-aware tool catalog
  • ✓arcade-mcp framework
  • ✓MCP Gateways and control plane
  • ✓Audit logging and evaluations
  • ✓Flexible deployment

Capabilities

✗text generation
✗image generation
✗video generation
✗code generation
✓workflow automation
✓api access
✗audio generation
✗fine tuning
✓agent orchestration

Use Cases

  • •Email and calendar assistant
  • •Slack and Teams workplace bots
  • •Governed internal MCP rollout
  • •Developer and DevOps agents

Ideal For

Best For

  • ✓Multi-user AI agents that must act in third-party SaaS with each end user's own OAuth permissions
  • ✓Enterprises that need agent actions enforced through their existing IdP and logged to a SIEM
  • ✓Teams standardising on MCP that want a governed gateway and registry for internal and remote MCP servers
  • ✓Regulated organisations that need agent tooling in a VPC or air-gapped environment

Not Ideal For

  • ✗Teams that need the widest possible integration catalog or data sync and webhooks. Competitor reviews describe Arcade as tool-calls-only, with a narrower first-party catalog than Composio, Nango or Paragon
  • ✗Buyers who require a fully open-source, inspectable runtime. The SDK and tools are open, but the runtime that executes tool calls is closed-source
  • ✗SaaS vendors that need per-tenant custom field mappings or customer-specific tool behaviour. Third-party comparisons say these are not built in

Market Analysis

Enterprise-gradeMCP-nativeDeveloper-first

Pros

  • ✓Solves the hardest production-agent problem: proving that an agent may perform an action on a resource for a specific user, using the existing IdP and OAuth
  • ✓Published, low-entry pricing with a no-card free tier and per-call metering
  • ✓Open-source MIT arcade-mcp framework and broad framework support, so teams avoid lock-in at the agent layer
  • ✓Enterprise deployment options (VPC, air-gapped, Helm) plus SOC 2, SSO and RBAC

Cons

  • ✗Narrower first-party catalog than rivals. Scalekit counts about 112 first-party integrations as of early 2026, and Merge notes many MCP servers are community-maintained, with variable quality
  • ✗Closed-source runtime with tool calls only: no data sync or webhooks, per Scalekit's comparison
  • ✗Org-level credential hierarchies and SIEM-exportable audit logs are described by a competitor as not yet comprehensive, and production evidence at enterprise scale is still accumulating
  • ✗Young company (founded 2024) that discloses few named customers. Most Hacker News activity is vendor-posted, with little independent practitioner discussion

Pricing

Free

$0

  • ✓2,000 auth events / month
  • ✓2,000 tool calls / month
  • ✓Fully managed on Arcade Cloud
  • ✓Community support on Discord and GitHub
  • ✓No credit card required

Team

From $25/mo

  • ✓$25/month platform fee plus usage
  • ✓$0.10 per auth event
  • ✓$0.01 per tool call
  • ✓Fully managed on Arcade Cloud
  • ✓Next-business-day email support

Enterprise

Contact for pricing

  • ✓Annual bundles of auth events and tool calls at discounted rates
  • ✓Arcade Cloud, customer VPC or fully air-gapped
  • ✓24/7 SLA-backed support
  • ✓Dedicated FDE and account team
  • ✓SSO, RBAC, audit logs and private registry access

List pricing is published. The Free tier includes 2,000 auth events and 2,000 tool calls a month. Team costs $25/month plus $0.10 per auth event and $0.01 per tool call. Enterprise is custom, sold as annual bundles of auth events and tool calls, and is the plan that includes VPC or air-gapped hosting, 24/7 SLA support, SSO, RBAC, audit logs and a private registry. Competitor Merge says Arcade's metering has hard-to-predict variables.

Security & Compliance

✓soc2
✗gdpr
✗hipaa
✗iso27001
✓sso
✗data residency

THE DAILY BRIEF

Enterprise AI insights for technology and business leaders, weekly.

beri.net

Subscribe at beri.net/subscribe for weekly AI insights delivered to your inbox.

LinkedIn: linkedin.com/in/rberi  |  X: x.com/rajeshberi

© 2026 Rajesh Beri. All rights reserved.

Arcade.dev is an authorization and tool-execution runtime for AI agents. It sits between agents and business systems such as Gmail, Slack, GitHub and Salesforce. It handles per-user OAuth, applies policy outside the model, and runs actions from a catalog of thousands of MCP tools. That lets engineering teams take agents from demo to production without building their own credential and permission plumbing.

Arcade.dev is an MCP-native runtime that controls what AI agents are allowed to do in enterprise systems. The vendor describes it as a single control point to enforce, execute and govern agent actions. On enforcement, Arcade connects to a company's identity provider and delegates authorization so an agent acts as the user it serves. It uses OAuth 2.0 with encrypted token storage, and policies are applied outside the model layer, where a prompt cannot override them. On execution, it offers a catalog the homepage puts at 8,000+ permission-aware tools; its docs and GitHub README cite 7,500+ tools across 81 MCP servers. Developers can add custom tools through the MIT-licensed arcade-mcp Python framework, where a tool declares the OAuth scopes it needs and Arcade handles the user consent prompt, token storage and refresh. Built-in evaluations aim to catch hallucinated or destructive tool calls. On governance, MCP Gateways and a central control plane grant and revoke tools, and audit logs can be streamed to SIEMs or emitted as OpenTelemetry. Arcade is framework-agnostic, with documented support for LangChain, OpenAI Agents, CrewAI, Google ADK, Vercel AI SDK, Mastra and others. It runs as Arcade Cloud, through the AWS and Azure marketplaces, self-hosted via Helm, in a customer VPC, or air-gapped. CEO Alex Salazar came from Okta and CTO Sam Partee from Redis. The company says it wrote the MCP tool-authorization specification. It raised a $12M seed round, then a $60M Series A in June 2026 led by SYN Ventures with Morgan Stanley and Wipro, for $72M in total. In August 2026 it acquired the MCP registry Smithery. Its main competitors are Composio, Merge Agent Handler, Nango and Scalekit.

Ideal Buyer

Platform or AI engineering leads who are moving multi-user agents into production. These agents must act inside SaaS systems on each user's behalf under existing identity and access policy.

Key Benefit

Agents can take real, audited actions in systems like Gmail, Slack, GitHub and Salesforce using each user's own delegated permissions. The team does not have to build OAuth, token storage or a policy layer.

At a Glance

Category
Agent Development
Pricing
Freemium, Usage-based, Contact for pricing
Target Market
CTOs, CIOs, Enterprise Developers, Security Teams, AI Platform Engineers
Deployment
Cloud-first, Self-hosted, Hybrid, API-based
Founded
2024
Headquarters
San Francisco, USA
Team Size
11-50

Key Features

  • ✓
    Delegated user authorization

    Connects to the company's identity provider and handles per-user OAuth 2.0 with encrypted token storage. Agents act with each user's own permissions, not a shared super-credential.

  • ✓
    Permission-aware tool catalog

    Prebuilt tools for Google Workspace, Slack, Teams, GitHub, Stripe, Snowflake, Postgres and more, cited as 7,500+ to 8,000+ across 81 MCP servers. This cuts integration build time.

  • ✓
    arcade-mcp framework

    An MIT-licensed Python framework for building custom MCP servers. Tools declare the OAuth scopes they need, and Arcade handles consent, token storage and refresh.

  • ✓
    MCP Gateways and control plane

    Central governance to grant or revoke tools, apply contextual and per-action policies, and register remote MCP servers alongside built-in integrations.

  • ✓
    Audit logging and evaluations

    Streams audit logs to SIEMs or emits OpenTelemetry. Built-in evals help catch hallucinated or destructive tool calls before they execute in production.

  • ✓
    Flexible deployment

    Runs on Arcade Cloud, the AWS and Azure marketplaces, self-hosted via Helm, in a customer VPC, or fully air-gapped, so it suits regulated environments.

Capabilities

✗text generation
✗image generation
✗video generation
✗code generation
✓workflow automation
✓api access
✗audio generation
✗fine tuning
✓agent orchestration

Use Cases

  • •
    Email and calendar assistant

    An agent reads and sends Gmail or Outlook messages and books meetings as the signed-in user. Each user grants consent once through Arcade's OAuth flow.

  • •
    Slack and Teams workplace bots

    Chat bots take OAuth-secured actions such as posting, summarising channels or creating tickets. Each action is checked against the requesting user's permissions and logged.

  • •
    Governed internal MCP rollout

    Platform teams publish internal and remote MCP servers through one gateway. They control which agents and users can call which tools, with audit logs sent to the SIEM.

  • •
    Developer and DevOps agents

    Coding agents act in GitHub, Vercel or Datadog with per-user delegated tokens. This avoids shared service accounts and keeps a traceable record of each change.

Ideal For

Best For

  • ✓Multi-user AI agents that must act in third-party SaaS with each end user's own OAuth permissions
  • ✓Enterprises that need agent actions enforced through their existing IdP and logged to a SIEM
  • ✓Teams standardising on MCP that want a governed gateway and registry for internal and remote MCP servers
  • ✓Regulated organisations that need agent tooling in a VPC or air-gapped environment

Not Ideal For

  • ✗Teams that need the widest possible integration catalog or data sync and webhooks. Competitor reviews describe Arcade as tool-calls-only, with a narrower first-party catalog than Composio, Nango or Paragon
  • ✗Buyers who require a fully open-source, inspectable runtime. The SDK and tools are open, but the runtime that executes tool calls is closed-source
  • ✗SaaS vendors that need per-tenant custom field mappings or customer-specific tool behaviour. Third-party comparisons say these are not built in

Integrations

✓SDK Available
SDK:PythonTypeScript

Deployment

✓On-Premise

Market Analysis

Enterprise-gradeMCP-nativeDeveloper-first

Pros

  • ✓Solves the hardest production-agent problem: proving that an agent may perform an action on a resource for a specific user, using the existing IdP and OAuth
  • ✓Published, low-entry pricing with a no-card free tier and per-call metering
  • ✓Open-source MIT arcade-mcp framework and broad framework support, so teams avoid lock-in at the agent layer
  • ✓Enterprise deployment options (VPC, air-gapped, Helm) plus SOC 2, SSO and RBAC

Cons

  • ✗Narrower first-party catalog than rivals. Scalekit counts about 112 first-party integrations as of early 2026, and Merge notes many MCP servers are community-maintained, with variable quality
  • ✗Closed-source runtime with tool calls only: no data sync or webhooks, per Scalekit's comparison
  • ✗Org-level credential hierarchies and SIEM-exportable audit logs are described by a competitor as not yet comprehensive, and production evidence at enterprise scale is still accumulating
  • ✗Young company (founded 2024) that discloses few named customers. Most Hacker News activity is vendor-posted, with little independent practitioner discussion

Pricing

Free

$0

  • ✓2,000 auth events / month
  • ✓2,000 tool calls / month
  • ✓Fully managed on Arcade Cloud
  • ✓Community support on Discord and GitHub
  • ✓No credit card required

Team

From $25/mo

  • ✓$25/month platform fee plus usage
  • ✓$0.10 per auth event
  • ✓$0.01 per tool call
  • ✓Fully managed on Arcade Cloud
  • ✓Next-business-day email support

Enterprise

Contact for pricing

  • ✓Annual bundles of auth events and tool calls at discounted rates
  • ✓Arcade Cloud, customer VPC or fully air-gapped
  • ✓24/7 SLA-backed support
  • ✓Dedicated FDE and account team
  • ✓SSO, RBAC, audit logs and private registry access

List pricing is published. The Free tier includes 2,000 auth events and 2,000 tool calls a month. Team costs $25/month plus $0.10 per auth event and $0.01 per tool call. Enterprise is custom, sold as annual bundles of auth events and tool calls, and is the plan that includes VPC or air-gapped hosting, 24/7 SLA support, SSO, RBAC, audit logs and a private registry. Competitor Merge says Arcade's metering has hard-to-predict variables.

Security & Compliance

✓soc2
✗gdpr
✗hipaa
✗iso27001
✓sso
✗data residency

Connect

Sources

This page was written from 10 sources, 7 on domains other than arcade.dev.

  1. 1.arcade.dev — arcade.devvendor
  2. 2.arcade.dev — pricingvendor
  3. 3.docs.arcade.dev — home
  4. 4.arcade.dev — smithery joins arcadevendor
  5. 5.github.com — arcade mcp
  6. 6.siliconangle.com — ai agent authorization startup arcade nabs 60m investment
  7. 7.thenextweb.com — arcade dev 60 million series a ai agent authorization
  8. 8.merge.dev — arcade alternatives
  9. 9.scalekit.com — arcade alternatives
  10. 10.hn.algolia.com — search
Newsletter

Stay Ahead of the Curve

Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.

Subscribe