Arcade
Also known as: Arcade AI
Actions runtime for AI agents, with delegated user authorization, more than 8,000 permission aware tools and MCP servers, inline policy hooks and a record of every action an agent takes.
Arcade calls itself the actions runtime for AI agents. It is one control point between agents and the systems they reach, which enforces policy, executes the action and records it. Its team comes from identity and data infrastructure companies including Okta, it says it authored the MCP tool authorization specification, and its core idea is that an agent should act as its user, never past that user's permissions, without the model ever holding a credential.
When an agent calls a tool, authentication runs against the customer's identity provider and authorization is delegated, so the agent acts within the user's scope and the user consents to each scope before a tool acts for them. An agent can also line up every authorization a job needs before it starts. Arcade offers more than 8,000 permission aware tools and connects custom or third party MCP servers, including hosted servers from Salesforce, ServiceNow, GitHub and Atlassian.
Tool Recommendation, added in October 2026, lets an agent describe a task and get back the few tools that fit instead of loading the whole catalog, which Arcade says cuts token costs by up to 37%. A Private Registry lets the people using agents see what a gateway offers and ask for what is missing, and the Arcade plugin installs into Claude Code, Cursor, Codex, VS Code, Copilot CLI and Claude Desktop.
Contextual access lets a customer run its own policy code inline at three points. These decide which tools a user can see, whether a call may run and with what inputs, and what output comes back, so sensitive actions can be blocked and personal data redacted. Each action leaves a record naming the agent, the user and the system, streamed to the customer's SIEM, and administrative changes land in a separate audit log with an API. An evaluation command scores how models choose and call tools across providers.
Arcade runs as a managed cloud, as a platform Arcade operates inside the customer's own Azure or AWS account, or self hosted with Helm, with VPC and air gapped options on Enterprise. It works with any model and framework the customer already uses. It does not run agents, orchestrate workflows or keep agent memory. Pricing is public. There is a free plan with 2,000 tool calls a month, Team at $25 a month plus $0.01 per tool call, and custom Enterprise.
From our comparison video
Arcade vs Composio: 6 Questions for Teams Whose Agents Act for Users
Our 6 minute video puts Arcade side by side with Composio. It covers what each one does and how each is priced, then when to choose which.
Recorded in October 2026. The profile here is live and updates as research is verified.
Vendor details
Canonical URL
https://arcade.dev
Category
Agent infrastructure
Subcategory
Agent auth and tool calling
Funding status
Independent; Arcade AI, Inc. Venture backed, with Cisco Investments joining its Series A in September 2026. Its team comes from Okta, Snowflake, Redis, Airbyte and MongoDB, per its homepage.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Arcade offers more than 8,000 permission aware tools plus custom and third party MCP servers. Hosted remote MCP servers cover Salesforce, ServiceNow, GitHub, Atlassian, Snowflake, Splunk, AWS and others, and OAuth auth providers span dozens of services. Identity sources include Okta, Microsoft Entra ID, Auth0 and Stytch. Guides cover LangChain, CrewAI, OpenAI Agents, Google ADK, Mastra, Vercel AI SDK and MCP clients such as Cursor, Claude and Copilot Studio.
In practice
You want your agent to act in Gmail and Salesforce for a user without the model touching OAuth tokens. Arcade authorizes the user against your identity provider and runs the tool within that user's scope, keeping credentials out of the prompt.
Your security team blocks the agent because no one can tell who authorized which action. You route tool calls through Arcade, which runs your policy hooks before and after each call and records every action with the agent, the user and the system for your SIEM.
You keep rebuilding OAuth and token handling for every new integration. You pull from Arcade's catalog of permission aware tools or bring your own MCP servers, and let Arcade handle the auth for each.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
7.5 / 14 capabilities · 54%
| Integrations & Tool Calling | Full |
|---|---|
|
More than 8,000 permission aware tools are available, plus custom or third party MCP servers that the customer builds or brings. Authorization is delegated, so the agent acts as its user within that user's scope, and Arcade handles OAuth across dozens of auth providers. Hosted remote MCP servers connect through Arcade to systems such as Salesforce, ServiceNow, GitHub, Atlassian and Snowflake. Tool Recommendation, added in October 2026, lets an agent describe a task and get back the few tools that fit instead of loading the whole catalog, which Arcade puts at up to 91% fewer input tokens per session, and a Private Registry lets users browse the tools behind a gateway and request missing ones. Sourcearcade.devread 2026-10-24 |
|
| Workflow Orchestration | Not documented |
|
Arcade executes the tool calls an agent makes and returns the results. Sequencing, branching and handoffs between agents live in the customer's own framework. Setup guides cover frameworks such as LangChain, CrewAI, OpenAI Agents, Google ADK and Mastra, among others. No workflow runtime of Arcade's own is named, so retries, routing and deterministic workflow steps sit outside the platform. Sourcearcade.devread 2026-10-24 |
|
| Knowledge Grounding & RAG | Not documented |
|
Agents get authenticated tools that read from and act in outside systems. Arcade names no index, graph, embeddings layer or other retrieval structure that it maintains over the customer's documents. Sourcearcade.devread 2026-10-24 |
|
| Human Oversight & Guardrails | Partial |
|
Contextual access runs the customer's own policy logic inline. An access hook decides which tools each user can see. A pre execution hook allows, denies or modifies inputs before each tool call, and a post execution hook allows, denies or modifies the output, so sensitive actions can be blocked and personal data redacted. Users authorize each scope before a tool acts for them, an agent can line up every authorization a job needs before it starts, and rate limits apply. A tool author can ask the user for input during a call through MCP elicitation, but Arcade names no step where a person reviews or approves an action before it commits. Sourcedocs.arcade.dev/en/operate/governance/contextual-accessread 2026-10-24 |
|
| Security, Identity & Governance | Full |
|
Authentication runs against the customer's identity provider, with Okta, Microsoft Entra ID, Auth0 and Stytch as user sources. Authorization is delegated, so an agent acts as its user and never past that user's scope. Contextual access hooks enforce the customer's policies on every tool call. Role based access control has organization admin, project admin and member roles, and SSO and audit logs are included on Enterprise. No security attestation is published. Sourcearcade.devread 2026-10-24 |
|
| Observability & Auditability | Full |
|
Every action taken through Arcade leaves one record naming the agent, the user and the system, streamed to the customer's SIEM, and execution tool logs show which tools ran, who they ran for and whether they succeeded. Administrative actions go to a separate audit log that is on by default, with OpenTelemetry export, a filterable dashboard and a REST API for pulling events into the customer's SIEM or compliance tooling. Sourcedocs.arcade.dev/en/operate/governance/audit-logsread 2026-10-24 |
|
| Memory & State Persistence | Not documented |
|
Each user's OAuth tokens and authorization grants are kept, so tools can act without asking for consent again. Those stored credentials carry no agent state. Arcade names no memory that keeps an agent's state or a user's context across runs. Sourcearcade.devread 2026-10-24 |
|
| Deployment & Data Residency | Full |
|
Customers choose between Arcade Cloud and their own infrastructure. A turnkey platform that Arcade operates can be deployed into the customer's cloud account, as an Azure Marketplace managed app or through an AWS private offer. A self managed Helm deployment can run on the customer's Kubernetes cluster, for data residency, network isolation or compliance. Enterprise adds VPC and fully air gapped deployment. Sourcedocs.arcade.dev/en/operate/deployread 2026-10-24 |
|
| Prebuilt Agents, Templates & Packs | Not documented |
|
The catalog holds tools and MCP servers that give an agent actions to take. The sample apps are example code. Arcade names no ready made workflows, templates or prebuilt agents of its own. Sourcearcade.devread 2026-10-24 |
|
| Triggers & Channel Coverage | Not documented |
|
Tools in Arcade run when an agent or MCP client calls them. No schedule, event, webhook or inbound queue that starts an agent run is named. Sourcearcade.devread 2026-10-24 |
|
| Model Flexibility & Routing | Full |
|
Any client, model or framework the customer picks can use Arcade's tools. Setup guides cover the customer's own LLM and frameworks such as LangChain, OpenAI Agents, Google ADK, CrewAI, Mastra and Vercel AI SDK, among others. The evaluation command runs against whichever providers and models the customer names. Sourcearcade.devread 2026-10-24 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Developers get an API, described in the Arcade API Reference, and a Contextual Access Webhook API. SDKs appear throughout the framework guides, alongside the Arcade CLI and an open source MCP framework in Python and JavaScript for building MCP servers that Arcade hosts or connects. The Arcade plugin, released in October 2026, installs into Claude Code, Cursor, Codex, VS Code, Copilot CLI and Claude Desktop, and any remote MCP client can connect through one server URL. The REST API behind audit logs is called with an Arcade API key. Sourcedocs.arcade.dev/llms.txtread 2026-10-24 |
|
| Testing, Debugging & Optimization | Full |
|
The arcade evals command runs evaluation suites that the customer writes against its tools. It gives detailed results and critic feedback, a failures only view, multiple runs for stability and comparisons across providers and models in one run. Capture mode records the tool calls a model makes without scoring them, so they can become test expectations, and Arcade describes its evals as catching hallucinated and destructive calls before they run. Sourcedocs.arcade.dev/en/build/create-tools/evaluate-tools/run-evaluationsread 2026-10-24 |
|
| Browser & Computer Use | Not documented |
|
Actions go through authenticated APIs and MCP tools. Arcade names no browser, desktop or remote computer session that it runs for an agent. Some toolkits connect to scraping and sandbox services that other vendors run. Sourcearcade.devread 2026-10-24 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Arcade now supports MCP Apps, so a tool can show an interactive interface inside the chat instead of a block of text. Drafts from Gmail, Outlook Mail and Slack open as editable messages that can be sent from the chat, Slack messages can be scheduled, and files can be uploaded and downloaded. Each app shows only the actions the user's permissions allow, so a toolset with no send tool shows no Send button.
Bears on: Human approval / guardrails
View sourceTool Recommendation, new in Arcade, matches each task to the few tools an agent needs at that moment instead of loading Arcade's whole catalog of more than 8,000 tools into context. It is built into the Arcade actions runtime, so governance and policy enforcement still apply to every recommended call.
Bears on: Agent capability
View sourceArcade announced a new Snowflake toolkit that lets agents discover warehouses, databases, schemas, tables, and columns, then run queries using the requesting user's identity. Existing Snowflake grants determine what the agent can access. The toolkit permits read only queries and blocks writes, schema changes, and role switches.
Bears on: MCP / tool calling / API
View sourcePricing
Free tier, Team $25 a month plus $0.01 per tool call, Enterprise custom
Monthly platform fee plus usage: tool calls
Included quota
Free includes 2,000 tool calls a month on Arcade Cloud with community support. Team adds no allowance of its own: every tool call costs $0.01 on top of the $25 monthly platform fee, with next business day email support.
What is public
Arcade publishes its rate card: a free plan with a monthly allowance, Team at $25 a month plus a per call rate, and custom Enterprise with deployment options and governance features.
Billing mechanics
A monthly platform fee on Team plus metered usage, where a tool call is one action an agent takes. Free includes a monthly allowance on Arcade Cloud. Enterprise buys annual bundles with discounted additional rates.
Cost watchouts
Tool calls add $0.01 each on Team, so cost follows how many actions agents take. VPC, air-gapped deployment, SSO and audit logs are Enterprise.
Variable cost rationale
One unit is metered, the tool call at $0.01, on top of a flat $25 a month on Team. Cost scales with the number of actions agents take. Enterprise buys annual bundles at discounted rates.
Additional watchouts
Every action an agent takes is a $0.01 tool call on Team, so busy agents move the bill. VPC and air gapped deployment, SSO, RBAC, audit logs and the private registry are Enterprise features.
Overage / add-ons
Team bills every tool call at $0.01 on top of the $25 platform fee; Enterprise buys annual bundles with discounted additional rates.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free plan: 2,000 tool calls a month, no credit card
Lowest paid plan
Team: $25 a month platform fee plus $0.01 per tool call
Commercial notes
Arcade is sold to engineering teams building production agents and to enterprises with strict security review. Its differentiator is auth: built by former Okta engineers, credentials never reach the model. Deploys to cloud, VPC, on premises, and locally.
Key ambiguities
Total cost depends on how many actions agents take; Enterprise bundle sizes and rates are quoted.
Cancellation / refund
Self serve plans are month to month and usage is billed in arrears. Cancellation details for paid plans are not publicly detailed.
Support SLA / resale
Community support on Discord and GitHub on Free, next business day email support on Team, and 24/7 SLA backed support with a dedicated forward deployed engineer and account team on Enterprise.
Missing data
Enterprise bundle sizes and discounted rates are quoted, not published.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…
Alternatives to Arcade
The closest documented capability profiles to Arcade among agent infrastructure platforms tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Portkey7.5 / 14Matches Arcade across all 14 documented capabilities
- Kong8.5 / 14Adds documented Knowledge Grounding & RAG
- Freeplay8.0 / 14Adds documented Triggers & Channel Coverage
- Metorial8.0 / 14Adds documented Triggers & Channel Coverage
- Stacklok8.0 / 14Adds documented Prebuilt Agents, Templates & PacksArcade vs Stacklok →
- Clawvisor6.5 / 14Fuller documented coverage on Human Oversight & Guardrails
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded