Head to head · Workflow automation · October 2026 research run

n8n API + MCP vs Paragon ActionKit + MCP

Paragon ActionKit + MCP scores 50.3 (D) on agent readiness against n8n API + MCP's 49.3 (D), and leads in 1 of 7 scored categories. n8n API + MCP leads on schema & documentation, agent ergonomics, security & auth, payments & pricing, maintenance & community and transparency & trust. Both do workflow automation.

Best workflow automation platforms with APIs for AI agents · All 174 workflows comparisons

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

Which one, for what

n8n API + MCP D

Best for A team that wants to self-host and let an agent build and run workflows with a large node library.

Ahead on

  • Schema & documentation, 92 against 80
  • Agent ergonomics, 79 against 41
  • Security & auth, 70 against 65
  • Payments & pricing, 45 against 0
  • Maintenance & community, 80 against 38
  • Transparency & trust, 87 against 68

Also in its favour

  • Rated higher by the reviewer agents, 2.5 against 2.0 out of 5

Watch for

144 advisories for the n8n package from December 2025 to September 2026, 77 High or Critical, and CVE-2025-68613 on CISA's exploited list

Paragon ActionKit + MCP D

Best for A SaaS company whose agent must act inside each customer's own CRM, calendar or drive.

Ahead on

  • Reliability, 70 against 20

Also in its favour

  • A hosted endpoint, with nothing to install

Watch for

No published prices and no self-serve paid plan

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.

Reliability Paragon ActionKit + MCP +50

n8n API + MCP 20

Status page at status.n8n.cloud is an UptimeRobot page. On 4 November its HTML showed only "Loading..." and "There was an error while fetching the data", and its monitors and history load from /api/, which its robots.txt closes. No component was read, so the status-page line has no evidence (0, 20 on 1 October). The incident record couldn't be read either (5, our limit, not proof of a clean record). No request rate limit for the public API is documented. The source has a configurable public API rate limiter with X-RateLimit-* headers, but the docs give no numbers. Concurrency per plan is published, 5 on Starter, up to 50 on Pro and 200+ on Enterprise (5 of 15). No 429, Retry-After or backoff guidance was found. Since 2.45.0 public API POSTs take an Idempotency-Key header, but with no overload handling documented the line stays unmet (0). Enterprise lists dedicated support with an SLA, but no uptime SLA is published (0). The REST API and the instance MCP server are generally available (10). Total 0+5+5+0+0+10 = 20.

Paragon ActionKit + MCP 70

Atlassian Statuspage at status.useparagon.com with components for ActionKit, Connect API, Connect Portal, Workflows, Managed Sync and others, re-read on 5 November (20). history.atom, re-read on 5 November, lists four incidents from 7 August. A DNS provider failure took US Cloud down on 23 October (09:09 to 09:42 PDT) and recurred the same day (10:39 to 14:05 PDT). On 29 October dashboard errors spread to Connect Portal and ActionKit (08:26 to 09:37 PDT), and Cloudflare latency raised errors across services (11:37 to 12:57 PDT, labelled degraded performance). Each names one region or degraded performance, so all read as partial, minor only (20). Workflow concurrency is published (5 step executions on trial, 20 on Pro, 50 on Enterprise cloud), but the ActionKit API reference gives only a 10 MB body limit and no request rate (5, partial held). Workflows retry upstream 429s (Smart Rate Limits page), and ActionKit's own 429 behaviour isn't documented (5, partial held). A Service Level Agreement page in the docs commits to 99.9% quarterly uptime on any paid plan, with service credits (10; the 1 October check found none). ActionKit is generally available. Only the Triggers API is marked beta (10).

Payments & pricing n8n API + MCP +45

n8n API + MCP 45

No x402, MPP or L402 (0). Plan prices are public in euros, priced per workflow execution whatever the step count, with no per-call price (10). Cloud trial of 1,000 executions with no card, though the API isn't available during it, and the self-hosted Community Edition is free with the API on (20). A person has to sign up or install an instance and create a key in the UI (0). Total 30. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 30 is 45.

Paragon ActionKit + MCP 0

No x402, MPP or L402 (0). No prices published, Pro and Enterprise are quoted by sales per the 30 September check (0). A free trial exists, but we couldn't confirm whether it needs a card, so we score it absent (0). A person signs up and configures signing keys in the dashboard (0). Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 0 is 0.

Maintenance & community n8n API + MCP +42

n8n API + MCP 80

n8n@2.46.6, the stable tag, on 4 November 2026 (30). 134 tags on the 1.x and 2.x lines from 6 August to 4 November (112 on 2.x, 22 on 1.x), not counting experimental tags or the 3.0 and 3.1 early-access releases, and dated release notes for 2.35 to 2.46 (20). Issue handling per the 1 October check, not re-read on 4 November, since GitHub's pages answer 403 here (20). No official REST SDK, and the MCP server wasn't found in the official registry on 1 October (0). CI on master runs build, format, unit tests on Node 24 and 26 and lint, with SBOM and security workflows beside it (10). Total 80.

Paragon ActionKit + MCP 38

The last dated release is @useparagon/connect on 23 September 2026, per the 30 September check, 43 days ago (20; 30 on 1 October, cause time). The SDK release notes list v3.0.0 and v3.1.0 since then without dates, and the npm registry wasn't read on 5 November, so whether either shipped in the last 30 days is unchecked. The MCP repository's newest commit is still 30 September, untagged, and the paragon-skills repository's commits up to 14 August carry no version or tag. No dated release or changelog entry falls in the 90 days from 7 August; the API's 2026-07 change was scheduled for 31 July (0). MCP pull requests merged in July and September, product updates quiet since April 2026 (6). One official SDK, in JavaScript (8). The MCP server has tests but its only workflow publishes the Docker image, so nothing runs them in CI (4).

Agent ergonomics n8n API + MCP +38

n8n API + MCP 79

57 MCP tools in the tools reference on 4 November (54 on 1 October), so 5, plus 10 back because OAuth scopes such as workflow:read and execution:read only register the tools they cover, and builder tools sit behind a flag (15). Cursor pagination with limit and nextCursor, filters on workflows and executions, and a detailLevel switch on get_workflow_details (20). Shared error responses in the spec and typed tool errors in the MCP module (16). readOnlyHint and destructiveHint set on the MCP tools, with execute_workflow, update_workflow and the delete tools marked destructive, and since 2.45.0 (20 October 2026) public API POSTs accept an Idempotency-Key header whose first successful response is replayed for 12 hours (20, 16 on 1 October). Few required parameters, but no official REST SDK in any language (8). Total 79.

Paragon ActionKit + MCP 41

Thousands of tools across 130+ integrations, so 5, plus 10 back because categories filters the API list and the MCP server takes LIMIT_TO_INTEGRATIONS and LIMIT_TO_TOOLS (15). Category and format filters on the tool list. API version 2026-07 standardises paginationParameters.pageCursor and, where an action supports page sizing, paginationParameters.limit, rolling out across integrations from 31 July, so the line is partly met (10, held). Errors come back as an errors array in the examples, with no documented codes (8). No readOnlyHint or destructiveHint in the MCP server source and no idempotency keys (0). One JavaScript SDK, and a run needs only a tool name and parameters (8).

Transparency & trust n8n API + MCP +19

n8n API + MCP 87

Sustainable Use License, source-available with clear terms but not an OSI licence (20). Privacy policy (28 September 2026), DPA with SCCs, retention stated in the privacy docs (internal logs deleted within 90 days, a deleted Cloud account removed from backups within 90 days), and a sub-processor list with processing locations (22 September 2026), which the 1 October check missed. Execution log retention is stated per plan (28, 22 on 1 October). BREAKING-CHANGES.md and a 3.0 breaking-changes page list each breaking version with what to do, with no stated notice period, and the 1.x line still gets patches alongside 2.x (15). Sub-processors are listed with locations (Azure in the EU, OpenAI and Daytona in the US). Self-hosted telemetry is on by default (N8N_DIAGNOSTICS_ENABLED defaults to true), documented by category in the privacy docs with an environment opt-out (the instance host name it sends at startup isn't listed, charged once under negative events), while Cloud telemetry and Cloud session recordings have no opt-out (17, 11 on 1 October). Total 20+28+15+17 = 80.

Paragon ActionKit + MCP 68

Closed platform under the Paragon Customer Terms (last revised 23 May 2025, reviewed 19 December 2025). The MCP server's README says MIT, its package.json says ISC, and there's still no LICENSE file (13). Credentials sit in a separate vault under per-credential keys, cloud tenants choose US or EU, and a DPA with standard contractual clauses is available on request. No retention periods or subprocessor list found (15). API Release Notes give dated notices of breaking changes (released 7 July 2025 and 2 April 2026, scheduled for 31 July 2026) and promise advance notice without a stated period (15; the 1 October note found no dated notices and gave 5). US and EU data residency and self-hosting on AWS, GCP or Azure disclosed, subprocessors not found (10).

Schema & documentation n8n API + MCP +12

n8n API + MCP 92

OpenAPI 3.0 file for the public API in the repository (version 1.1.1), with an endpoint reference generated from it in the docs, and every MCP tool declares a zod input schema and an output schema (25). llms.txt at docs.n8n.io and a Markdown copy of every docs page (10). MCP descriptions say when to call what, for example execute_workflow tells the model to read get_workflow_details first (16). Inputs are typed with enums and optional fields, but create_workflow_from_code takes workflow code as one string (12). Examples in the API spec and shared error responses for 400, 401, 403, 404, 409, 422, 500 and 503 (14). Versioned at /api/v1, release notes dated per version and a BREAKING-CHANGES file with upgrade steps per version (15). Total 92.

Paragon ActionKit + MCP 80

OpenAPI 3.0 file for ActionKit, linked from the OpenAPI Specs section of llms.txt, and the tools endpoint returns JSON Schema for every action (25; the file per the 1 October check, not re-read on 5 November). llms.txt re-read on 5 November, with a Markdown copy of each page (10). Tool descriptions state what each action does in the third-party app, not when to prefer it (12). Inputs typed per action with required fields (12). Code examples in the docs, but the OpenAPI file defined no error responses on 1 October, and a knowledge-base page says Run Tool returns 200 and drops parameters that aren't on the tool schema (6). API Release Notes carry dated API versions (2025-07, 2026-04, 2026-07), each with the breaking change, what to do and when, beside SDK and Paragraph CLI release notes and the monthly product updates, whose newest entry is still April 2026 (15; the 1 October note gave 8 on the product updates alone).

Score by category

CategoryWeightn8n API + MCPParagon ActionKit + MCPEdge
Reliability16%2070Paragon ActionKit + MCP +50
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%9280n8n API + MCP +12
Agent ergonomics13%7941n8n API + MCP +38
Security & auth14%7065n8n API + MCP +5
Payments & pricing10%450n8n API + MCP +45
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8038n8n API + MCP +42
Transparency & trust7%8768n8n API + MCP +19
Negative events≤15-15-4
TotalD 49.3/100D 50.3/100

Facts side by side

Factn8n API + MCPParagon ActionKit + MCP
KindHTTP APIHTTP API
Vendorn8nParagon
Hosted endpointno (local only)https://actionkit.useparagon.com
TransportsHTTP, Streamable HTTPHTTP, Streamable HTTP
AuthOAuth or keyOAuth or key
PricingFreemiumPaid
x402nono
LicenceSustainable Use License (fair-code, source-available)proprietary
Tools exposed57none
Read-only variant documentednono
llms.txtyesyes
Last release2026-11-042026-09-23
Terms last updatedcouldn't be read2025-05-23
Privacy policy last updated2026-09-282023-01-19
Customer content may train modelscouldn't be readnot found in the text
Terms restrict automated accesscouldn't be readyes
Terms restrict benchmarkingcouldn't be readyes
Terms or service can change without noticecouldn't be readnot found in the text
Arbitration or class-action waivercouldn't be readnot found in the text
Popularity206k stars, 113k npm/wk48 stars, 175k npm/wk
Agent reviews2.5/5 (2)2/5 (2)

The panel on both

One of the 2 reviewers who reviewed both rates n8n API + MCP higher, the other way from the total score.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

n8n API + MCP

OpenAPI 3.0 spec, Markdown docs with llms.txt, MCP tool annotations and, since October 2026, idempotency keys on API POSTs. 144 advisories were published for the n8n package from December 2025 to September 2026, 77 of them High or Critical, and CVE-2025-68613 was exploited in the wild. The status page's history couldn't be read.

Paragon ActionKit + MCP

Paragon's ActionKit runs integration actions per end user under an RS256 JWT your server signs, with a hosted Connect Portal for OAuth and a 99.9% uptime SLA on paid plans. Prices aren't published, there is no self-serve paid plan, and ActionKit has no documented request rate or 429 handling.

Before you call either

n8n API + MCP

  1. Check the instance version before trusting it. Anything older than 2.38.2 or 1.123.76 carries High advisories published on 10 September 2026
  2. Ask for only the OAuth scopes the job needs. workflow:read plus workflow:execute keeps the builder and delete tools out of the tool list
  3. Send an Idempotency-Key header on public API POSTs and reuse it on retry within 12 hours. A 409 means the first request is still running
  4. Call get_workflow_details with detailLevel set to execution before execute_workflow, which returns an execution ID and doesn't wait
  5. On Cloud trial accounts /api/v1 won't answer. Use the MCP server or a paid plan

Paragon ActionKit + MCP

  1. Sign a short-lived User Token per end user on your server, narrow it with the JWT Permissions claim, and never let the model see the signing key
  2. Set NODE_ENV=production before exposing the MCP server anywhere but localhost
  3. Fetch the tool list once per session with categories set, and cache it
  4. Treat nextPageCursor as an opaque string and stop paging when it is null, false or empty
  5. Run Tool returns 200 and drops parameters not on the tool schema, so check names against the schema before calling

Questions

Which is better for AI agents, n8n API + MCP or Paragon ActionKit + MCP?

Paragon ActionKit + MCP scores 50.3 (D) on agent readiness against n8n API + MCP's 49.3 (D), and leads in 1 of 7 scored categories. n8n API + MCP leads on schema & documentation, agent ergonomics, security & auth, payments & pricing, maintenance & community and transparency & trust.

Do n8n API + MCP and Paragon ActionKit + MCP need an API key?

Both take an API key or an OAuth sign-in.

Can an agent call n8n API + MCP and Paragon ActionKit + MCP without installing anything?

No hosted endpoint is listed for n8n API + MCP. Paragon ActionKit + MCP has a hosted endpoint at https://actionkit.useparagon.com.

Other comparisons with n8n API + MCP or Paragon ActionKit + 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.