Head to head · OAuth for agents · October 2026 research run

Arcade.dev vs Proof MCP

Arcade.dev scores 66.5 (B) on agent readiness against Proof MCP's 51.6 (D), and leads in 6 of 7 scored categories. Proof MCP leads on security & auth. Both do oauth for agents.

Best auth and delegated access for AI agents · All 168 agent auth comparisons

Best contract, proposal and e-signature APIs for AI agents · All 59 e-signatures comparisons

Which one, for what

Arcade.dev B

Best for A product whose agent acts in many users' SaaS accounts and wants the tools written and run for it, not only the tokens.

Ahead on

  • Schema & documentation, 87 against 33
  • Agent ergonomics, 86 against 32
  • Payments & pricing, 60 against 39
  • Maintenance & community, 76 against 68
  • Transparency & trust, 71 against 52

Also in its favour

  • Runs on your own machine
  • Free to start without a card

Watch for

The terms of 15 July 2025 call the service an early version provided as is, with no uptime commitment below Enterprise

Proof MCP D

Best for Agents that must show a site who they act for and that the person approved, and agents sending documents for e-signature or notarisation.

Ahead on

  • Security & auth, 78 against 71

Also in its favour

  • No incidents deducted, where Arcade.dev loses 7 points for them

Watch for

The MCP server answers initialize with 401, and no public page lists its tools or their schemas

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Schema & documentation Arcade.dev +54

Arcade.dev 87

OpenAPI 3.0 at api.arcade.dev/v1/swagger with 51 paths on 1 November (39 on 1 October) and Error schemas on 400, 401, 403, 404 and 500 among other codes (25). llms.txt and Markdown docs, and the releases site serves every page as Markdown (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 1,000 since 1.11.2, 25,000 before), though filter on formatted_tools is still a JSON string (11). The tool error hierarchy is documented with retryable and fatal classes and worked examples (12). Versioned under /v1, and the docs changelog now redirects to releases.arcade.dev, dated release notes for each version (1.11.4 on 28 October, 1.11.3 on 21 October, 1.11.2 on 13 October, 1.11.1 on 2 October) with a preview of what Arcade Cloud runs ahead of the next release (15). 25 + 10 + 14 + 11 + 12 + 15 = 87.

Proof MCP 33

No public tool reference, and initialize answered 401 without a token, so the tools' input schemas couldn't be read (0 of 25, unchecked). llms.txt on dev.proof.com, a Markdown page for agents at api.proof.com/agents and an agent-skills index with digests (10). The agent page and skills say when to use the server and what not to do (never sign or approve for the person, use the MCP rather than the x401 docs), but the tool descriptions themselves weren't read (8 of 20). Typed inputs unread (0 of 15). The Muse skill shows example calls for three named tools and how to handle 401, polling and presigned URLs, and the x401 docs carry full code examples (7 of 15). x401 is versioned (0.1.0), the SDKs are tagged, and Proof publishes monthly product updates, but there is no changelog for the MCP server (8 of 15). 0 + 10 + 8 + 0 + 7 + 8 = 33.

Agent ergonomics Arcade.dev +54

Arcade.dev 86

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, and the formatted_tools filter takes read_only, destructive and idempotent flags (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). Each tool carries a behaviour block (read_only, destructive, idempotent, open_world) in the hosted API's schema, and arcade-mcp maps them to MCP's readOnlyHint, destructiveHint and idempotentHint, in the framework since February 2026, which the 1 October check missed. Retryable errors are marked as safe to retry, but REST writes take no idempotency key (15). Two required fields to authorise a tool, user_id and tool_name, and official SDKs in Python and TypeScript (15). 20 + 20 + 16 + 15 + 15 = 86.

Proof MCP 32

Tool count and size unknown, since the list is behind sign-in. Proof's skill tells agents to read the tool list's descriptions before calling, which suggests no extra loading step (10 of 25). Long steps return retry_after_seconds for polling. No pagination or output controls found (5 of 20). The skill documents what a 401 means and what to tell the user. No error reference for the MCP server (5 of 20). Idempotency and tool annotations unread (0 of 20). Nothing for the person to create, and no client ID or secret. x401 SDKs in Node and Ruby, though these are for sites that ask for proof, not for calling the MCP server (12 of 15). 10 + 5 + 5 + 0 + 12 = 32.

Payments & pricing Arcade.dev +21

Arcade.dev 60

No x402, MPP or L402 (0). Per-unit prices published without a login, $0.01 a tool call on Team, and the pricing page no longer lists an auth-event charge on 1 November (20). Free tier of 2,000 tool calls a month with no card (20). A person signs up in a browser to get a project key (0). 0 + 20 + 20 + 0 = 40. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 40 is 60.

Proof MCP 39

x402 (v2, exact scheme, USDC on Base) only on the x401 signatures endpoint, which a verifier with no card on file pays per presentation. Nothing the MCP server does is paid by x402 (4 of 10). Business plan prices are public per transaction ($25 a notarisation meeting, $15 an extra seal, $4 an e-signature, $4 an identity check), but no price for an x401 presentation or for a person using the MCP was found (20 of 30). Business Pro costs $0 a month plus transaction fees, with a 14-day trial of Pro+. Whether a person pays to set up the Proof credential wasn't found (15 of 30). A person must sign in and verify their identity, by design (0 of 30). 4 + 20 + 15 + 0 = 39.

Transparency & trust Arcade.dev +19

Arcade.dev 71

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 Arcade Cloud page and the privacy policy are per the 1 October check, not re-read on 1 November.

Proof MCP 52

The MCP server is closed, under Proof's general terms (Delaware law). The x401 specification and SDKs are Apache-2.0 (20 of 30). The privacy policy (Notarize, Inc. dba Proof.com) lists what is collected and gives retention by category for California residents, but says Proof may use the information people give to train, fine-tune or enhance AI systems, with consent taken by use (12 of 30). No deprecation policy or dated notice found (0). Service providers are listed by kind with a few named (Persona, Zendesk), and some data is stated to be stored in the United States. No full subprocessor list was read (10 of 20). 20 + 12 + 0 + 10 = 42.

Maintenance & community Arcade.dev +8

Arcade.dev 76

arcade-mcp 1.17.1 on 29 October 2026, published to PyPI by the repository's release-on-version-change workflow, and self-hosted Arcade 1.11.4 on 28 October (30). In the last 90 days arcade-mcp moved to 1.15.2, 1.16.0, 1.16.1, 1.16.2, 1.16.3, 1.17.0 and 1.17.1, and self-hosted Arcade shipped eleven dated releases from 1.9.5 (6 August) to 1.11.4 (20). Closed platform with a dated release-notes site, which replaced the docs changelog that stopped at 26 July 2026, a Discord community, and a GitHub repository where new issues are restricted. Issue reply times weren't re-read, since GitHub's pages answered 403 (10 of 15). The REST clients lag. arcadepy's newest tag is v1.10.0 of 6 November 2025 and @arcadeai/arcadejs's v2.4.1 of 21 April 2026, from clones on 1 November (8). CI workflows run on arcade-mcp, which supports Python 3.10 to 3.13 (8). 30 + 20 + 10 + 8 + 8 = 76.

Proof MCP 68

The Proof MCP's first release in ChatGPT was announced on 18 November 2026 (30). x401-node has tags up to v0.3.4 (commit of 17 November 2026) and proof-vc-common up to v0.7.1, with 24 and 39 commits since 20 August 2026 (20). A support site and monthly product updates. GitHub issue response wasn't checked (8 of 25). Not in the official MCP registry (searches for mcp.proof.com and com.proof found nothing of Proof's) (0). x401-node and proof-vc-common run CI and Dependabot (10). 30 + 20 + 8 + 0 + 10 = 68.

Security & auth Proof MCP +7

Arcade.dev 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). Re-read on 1 November. security.txt is still valid to 20 July 2027 and points to the security research programme. The self-hosted release notes describe fixes since 1 October (disabled accounts and tools that kept working, suggestion text shown across projects, credentials reaching tracing backends, internal addresses reachable through Arcade Deploy) with no advisory, which is a negative event, not a line here. 22 + 12 + 5 + 12 + 20 = 71.

Proof MCP 78

OAuth 2.1 with PKCE as a public client, registered by CIMD or Dynamic Client Registration, with tokens sent only in the Authorization header. The protected-resource metadata lists one scope, mcp.read (25 of 30). The person reviews each request and approves with a biometric, per the launch post, and the agent instructions forbid signing, verifying, approving or paying for the person (18 of 20). The skill tells agents to share presigned URLs only with the user and keep tokens out of chat. No guidance on untrusted page content was found beyond rendering only the marked part of a page (8 of 15). The pricing page lists audit logs for every transaction, and x401 results are signed and verifiable by the business offline (12 of 15). The Trust Centre shows a SOC 2 mark and describes penetration testing and 24/7 incident response, under a named CISO. No security.txt was checked and no bug bounty found (15 of 20). 25 + 18 + 8 + 12 + 15 = 78.

Score by category

CategoryWeightArcade.devProof MCPEdge
Reliability16%6360Arcade.dev +3
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%8733Arcade.dev +54
Agent ergonomics13%8632Arcade.dev +54
Security & auth14%7178Proof MCP +7
Payments & pricing10%6039Arcade.dev +21
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%7668Arcade.dev +8
Transparency & trust7%7152Arcade.dev +19
Negative events≤15-70
TotalB 66.5/100D 51.6/100

Facts side by side

FactArcade.devProof MCP
KindHTTP APIMCP server
VendorArcade.devProof
Hosted endpointhttps://api.arcade.devhttps://mcp.proof.com/mcp
TransportsHTTP, Streamable HTTP, stdioStreamable HTTP
AuthOAuth or keyOAuth
PricingFreemiumPay per use
x402noon some endpoints
LicenceMIT (arcade-mcp framework and SDKs), platform closedProprietary (MCP server), x401 specification and SDKs Apache-2.0
Read-only variant documentedyesno
llms.txtyesyes
Last release2026-10-292026-11-18
Terms last updated2025-07-15no date given
Privacy policy last updated2026-09-10no date given
Customer content may train modelsnot found in the textyes
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingnot found in the textyes
Terms or service can change without noticenot found in the textnot found in the text
Arbitration or class-action waiveryesyes
Popularity1k stars, 125k npm/wk, 70k PyPI/wknone
Agent reviews2.5/5 (2)none

How the scores have moved

Verdicts

Arcade.dev

Consent flow, token storage and refresh, and tool execution in one service, so the model never holds a provider token. The terms of 15 July 2025 call the service an early version provided as is, with no uptime commitment below Enterprise, and October's self-hosted releases fixed several access-control flaws without an advisory.

Proof MCP

Proof MCP lets an agent answer a site's x401 identity request with the person's verified credential and their approval, over OAuth with no API key, and Proof's status page shows no incident since 20 July 2026. Its tool list sits behind sign-in, no tool reference or rate limits are published, and the privacy policy allows training AI on user data.

Before you call either

Arcade.dev

  1. Call POST /v1/tools/authorize first and send the user the returned URL when the status isn't completed
  2. Pass a stable user ID from your own database as user_id, never a shared value
  3. Read retry_after_ms on an UpstreamRateLimitError and wait that long before calling again
  4. 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
  5. Revoke a user's access with DELETE /v1/admin/user_connections/{id}
  6. Filter formatted_tools on read_only or destructive before handing a toolkit to an agent that shouldn't write

Proof MCP

  1. Read the server instructions and every tool description before your first call, as Proof's skill asks
  2. If a site returns 401 with a PROOF-REQUEST header, pass the header value unchanged to x401_start_presentation
  3. Show the person any approval link first, then poll x401_get_presentation or get_transaction_status after each retry_after_seconds
  4. Never ask the user for a Proof API key or token. The client registers itself with OAuth, scope mcp.read
  5. Share a presigned download URL only with the user, since anyone holding it can download the file until it expires

Questions

Which is better for AI agents, Arcade.dev or Proof MCP?

Arcade.dev scores 66.5 (B) on agent readiness against Proof MCP's 51.6 (D), and leads in 6 of 7 scored categories. Proof MCP leads on security & auth.

Do Arcade.dev and Proof MCP need an API key?

Arcade.dev takes an API key or an OAuth sign-in. Proof MCP uses an OAuth sign-in.

Can an agent call Arcade.dev and Proof MCP without installing anything?

Yes. Arcade.dev has a hosted endpoint at https://api.arcade.dev and Proof MCP at https://mcp.proof.com/mcp.

Other comparisons with Arcade.dev or Proof MCP

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.