Arcade.dev by Arcade.dev

# Fix list: Arcade.dev

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/arcade, the October 2026 research run, assessed 1 October 2026. Grade B, 67 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Arcade.dev: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Payments & pricing, 40 out of 100, up to 7.5 more on the total

Why it scored 40: No x402, MPP or L402 (0). Per-unit prices published without a login, $0.01 a tool call and $0.10 an auth event on Team (20). Free tier of 2,000 auth events and 2,000 tool calls a month with no card (20). A person signs up in a browser to get a project key (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 2. Reliability, 63 out of 100, up to 7.4 more on the total

Why it scored 63: Atlassian Statuspage at status.arcade.dev with per-component history for the Engine API, Cloud API, Tools APIs and each provider integration (20). One incident in the last 90 days, user registration and plan upgrades on 17 July 2026 from 15:37 to 17:59 UTC, which didn't touch tool calls (20). The only rate limit with a number is 100 requests a minute on the audit-log API, nothing for the authorise or execute calls (5). Upstream rate limits come back as UpstreamRateLimitError with retry_after_ms and a documented wait-and-retry rule, but the OpenAPI file defines no 429 and no idempotency key (8). Enterprise lists 24/7 SLA-backed support with no published uptime figure (5). The terms dated 15 July 2025 still call the service an early version under development, so half marks for availability (5).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 3. Security & auth, 71 out of 100, up to 5.1 more on the total

Why it scored 71: Gateways sign end users in with OAuth through Arcade Auth or your own OIDC provider, and provider tokens are requested per tool scope, but the REST fallback is one project API key plus an `Arcade-User-ID` header that can name any user (22). Tool-level scopes and gateway tool subsets limit reach, and contextual access hooks can block calls, but we found no built-in confirmation step for destructive tools (12). Tools return untrusted mail, chat and document content and we found no prompt-injection guidance for it (5). Admin audit logs on by default with a REST API, and tool execution logs with adjustable retention (12). security.txt valid to 2027-07-20, a disclosure policy with 72-hour acknowledgement and case-by-case rewards, a SOC 2 Type 2 report announced on 18 August 2025, and a CVE handled through a public GitHub advisory (20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 4. Agent ergonomics, 79 out of 100, up to 3.4 more on the total

Why it scored 79: Gateways expose a chosen subset of tools, and formatted_tools filters by toolkit with limit and offset, so context can be sized (20). limit and offset with user_id and provider_id filters on list endpoints (20). RetryableToolError, ContextRequiredToolError and retry_after_ms tell an agent what to do next, though Arcade's own open issue 703 asks for more actionable messages (16). Retryable errors are marked as safe to retry, but there are no idempotency keys and we didn't find readOnlyHint or destructiveHint on the hosted tools (8). Two required fields to authorise a tool, user_id and tool_name, and official SDKs in Python and TypeScript (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 5. Schema & documentation, 82 out of 100, up to 2.9 more on the total

Why it scored 82: OpenAPI 3.0 at api.arcade.dev/v1/swagger with 39 paths and Error schemas on 400, 401, 404 and 500 (25). llms.txt and Markdown docs (10). Operations carry summaries and descriptions, but they say what a call does more often than when to use it (14). Required fields and limits are typed (limit max 25,000), though `filter` on formatted_tools is a JSON string (11). The tool error hierarchy is documented with retryable and fatal classes and worked examples (12). Versioned under /v1 and a weekly changelog, but its latest entry is 26 July 2026 (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 6. Transparency & trust, 72 out of 100, up to 2.5 more on the total

Made of editorial 47, provenance 96.

Why it scored 72: Closed platform under terms that name Arcade AI, Inc. and California law, with the MIT arcade-mcp framework beside it (15). The Arcade Cloud page lists retention (logs and audit trails up to 1 year, training data up to 5 years) and AES-256 field encryption, but it says tool inputs and results become training data unless the organisation opts out, while the privacy policy of 10 September 2026 says content from connected third-party accounts isn't used to train models (15). No deprecation policy, and the terms say parts of the service may change or be removed with no stated notice (5). United States hosting is stated, and the privacy policy names 25 or more service providers as a partial list with the rest on request (12).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: arcade.dev, registered 2019-03-26 (7 years) (11 of 15)

## 7. Maintenance & community, 74 out of 100, up to 2.3 more on the total

Why it scored 74: arcade-mcp 1.16.1 on PyPI on 25 September 2026 (30). arcade-mcp shipped 1.15.1, 1.15.2, 1.16.0 and 1.16.1 in the last 90 days (20). Closed platform with a public changelog that stopped at 26 July 2026, a Discord community, and a GitHub repo where new issues are restricted and the oldest open bug is about a year old (8). The REST clients lag, arcadepy 1.10.0 dates from 6 November 2025 and @arcadeai/arcadejs 2.4.1 from April 2026 (8). CI workflows run on arcade-mcp, which supports Python 3.10 to 3.13 (8).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2025-12-02, CVE-2025-66454 (GHSA-g2jx-37x6-6438, CVSS 6.5). The HTTP worker in arcade-mcp shipped a hardcoded default worker secret, so anyone could forge a token and list or call every tool on a self-hosted worker set up by the official guide. Fixed in 1.9.1 and disclosed in public, so we deduct 2 of a possible 15 (https://github.com/ArcadeAI/arcade-mcp/security/advisories/GHSA-g2jx-37x6-6438)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- We couldn't find rate limits for `/v1/tools/authorize` or tool execution, or whether failed tool calls are billed.
- Whether the training-data clause on the Arcade Cloud page applies to results from connected accounts such as Gmail, given the privacy policy says that content isn't used for training.
- Whether the hosted tools set readOnlyHint and destructiveHint annotations.
- The trust centre at trust.arcade.dev didn't render for us, so we couldn't read the subprocessor list or the SOC 2 report scope.

## Weaknesses

- The terms of 15 July 2025 call the service an early version provided as is, with no uptime commitment below Enterprise
- Tool inputs and results are training data for up to 5 years unless the organisation opts out
- No published rate limits for the authorise or execute calls, and no 429 in the OpenAPI file
- The public changelog's latest entry is 26 July 2026 and arcadepy hasn't been released since 6 November 2025
- Hosting in the United States only outside Enterprise

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Call `POST /v1/tools/authorize` first and send the user the returned URL when the status isn't completed
- Pass a stable user ID from your own database as user_id, never a shared value
- Read retry_after_ms on an UpstreamRateLimitError and wait that long before calling again
- Register your own OAuth app and a custom user verifier before real users sign in, because the default apps only accept members of your Arcade project
- Revoke a user's access with `DELETE /v1/admin/user_connections/{id}`

## What the review panel asked for

- A keyless trial route
- per-user credentials on REST
- training off by default

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.