Head to head · Video generation · October 2026 research run

Magic Hour MCP vs Replicate video models

Replicate video models scores 64.2 (B) on agent readiness against Magic Hour MCP's 50.7 (D), and leads in 5 of 7 scored categories. Magic Hour MCP leads on maintenance & community. Both do video generation.

Best video generation for AI agents · All 183 video comparisons

Which one, for what

Magic Hour MCP D

Best for Agents making short-form video and images, from talking photos and ads to lip-synced translations, on a credit plan.

Ahead on

  • Maintenance & community, 75 against 68

Watch for

About 160 KB of tool definitions

Replicate video models B

Best for An agent that wants several vendors' video models behind one token and one request shape, or open models such as Wan next to closed ones.

Ahead on

  • Reliability, 75 against 35
  • Agent ergonomics, 72 against 55
  • Payments & pricing, 30 against 15
  • Transparency & trust, 76 against 58

Also in its favour

  • Runs on your own machine

Watch for

Tokens have no scopes, expiry or spend cap, and any token can run every model and delete account resources

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 Replicate video models +40

Magic Hour MCP 35

Status page at status.magichour.ai, all services online, with maintenance and previous incidents sections; no MCP component read (15). Incident history not read (5). No API rate limits found (0). No retry guidance found; wait_for_* tools poll long jobs (5 of 15). No SLA found (0). Not labelled beta (10). 15 + 5 + 0 + 5 + 0 + 10 = 35.

Replicate video models 75

Graded as a hosted service, on the predictions API. Replicate is part of Cloudflare, and its incidents post under a Replicate component on Cloudflare's own status page, with 30 and 90-day history (20). Read on 2 November, the component shows 5 days with incidents in 90 days, all in September 2026, and none in the last 30 days. Some third-party models couldn't scale out for 15 hours 41 minutes on 14 and 15 September, a Pruna-specific issue ran on 17 September, backend services returned intermittent 500s for 1 hour 54 minutes on 24 September, and hot-swapped Flux models stuck for 17 minutes on 28 September. Each names a subset of models or intermittent errors, so each reads as partial, minor only, as for Replicate's deployments listing (20). Limits of 600 prediction creates and 3,000 other requests a minute are published (15). A 429 body states when the limit resets and the error-code page gives retry advice per code, but no Retry-After header or idempotency key is documented (10 of 15). The enterprise page mentions SLAs with no published terms, and the terms of service disclaim service levels (0). The predictions API and official models are generally available (10). 20 + 20 + 15 + 10 + 0 + 10 = 75.

Transparency & trust Replicate video models +18

Magic Hour MCP 58

Closed service under terms naming Magic Hour AI, Inc., revised 3 October 2026 (15). The privacy policy says prompts, uploads and outputs aren't sold or used to train models, for free and paid accounts, and MCP requests are handled the same way; retention periods aren't given (20 of 30). No deprecation policy found (5 of 20). Subprocessors not read (10 of 20, model providers named in tool docs). 15 + 20 + 5 + 10 = 50.

Replicate video models 76

Closed platform under terms of service dated 1 April 2026 that name Replicate, LLC. The clients are Apache-2.0 and each model keeps its own licence, for example Apache 2.0 for Wan 2.7 (17). The docs say API inputs, outputs, files and logs are deleted after one hour and website predictions are kept until deleted. The privacy policy gives no retention period, and we found no statement on what the outside providers behind official models (Google, Kuaishou, Runway and others) keep (18). No written deprecation policy, but the changelog dates platform deprecations, such as the monthly spend limit deprecated on 1 July 2025 and language model training stopped on 17 December 2024, read as on Replicate's deployments listing. The terms say Replicate doesn't promise to keep any model and may discontinue parts of the service without notice, while the official-models page promises a stable interface (12, 6 on 8 October). A subprocessor page lists 17 companies, each located in the United States. The model providers aren't on it (15). 17 + 18 + 12 + 15 = 62.

Agent ergonomics Replicate video models +17

Magic Hour MCP 55

50 tools totalling about 160 KB of definitions (5). list_saved_items and project getters; limited size controls (10 of 20). Error handling not documented (5 of 20). Every tool annotated, 14 read-only and the three deletes marked destructive (20). Few required inputs and SDKs (Python, Node and others listed in the docs) (15). 5 + 10 + 5 + 20 + 15 = 55.

Replicate video models 72

A prediction returns a small JSON object with the output as a file URL. The MCP server exposes one tool per HTTP operation, 36 in the OpenAPI document, with an experimental local code mode that reduces them to two (16). Cursor pagination on list endpoints, webhook_events_filter, and Prefer: wait with a chosen duration (18). A 429 body with the wait time, coded errors with suggested fixes and an error field on failed predictions (15). No idempotency key. Failed runs aren't charged, and Cancel-After, a cancel endpoint and webhooks limit polling and runaway jobs. MCP annotations weren't checked (8). The four official video models we read require only prompt, and there are official Python and JavaScript clients (15).

Payments & pricing Replicate video models +15

Magic Hour MCP 15

No machine payment protocol (0). Plan and credit-pack prices public, per-tool credit costs not read (10). 400 sign-up credits with no card, but API access sits on paid plans (5 of 20). A person signs up for a key in a browser (0). Tool discovery is open, but calls need a key (0). 0 + 10 + 5 + 0 = 15.

Replicate video models 30

No x402, MPP or L402 (0). Per-second prices for each official video model are published on its model page without a login, and hardware prices on the pricing page (20). No standing free tier. The try-for-free collection lets a new account run a few models, minimax/video-01 among them, a limited number of times before buying credit (10 of 20). Signup and token creation are a browser flow (0).

Maintenance & community Magic Hour MCP +7

Magic Hour MCP 75

Changelog entries on 3, 4 and 5 November 2026 (30). More than three dated entries in 90 days (20). Changelog and support; issues on the public MCP repository not read (10 of 25). Registry entry io.github.magichourhq/magic-hour (0.1.1, 4 November) (15). No package health read (0). 30 + 20 + 10 + 15 + 0 = 75. Registry presence read from Anchor Terminal's own daily import of the official MCP registry (/api/v1/registry.json, 6 November 2026), not the registry's API.

Replicate video models 68

The brief counts the last published model for a closed service, and every dated source of the vendor's own. Kling 4.0 (kwaivgi/kling-v4.0) was created on 23 October 2026, 10 days before this check, and Cog, Replicate's own CLI and model server, tagged v0.23.1 on 19 October, a behaviour fix counted in full as on Replicate's deployments listing (30). In the 90 days from 4 August, Cog tagged v0.22.0 (13 August), v0.23.0 (22 September) and v0.23.1 (19 October), and alibaba/wan-3 (15 September) and Kling 4.0 (23 October) were published, five dated releases. The public changelog's newest entry is still 21 April 2026 (20, 0 on 8 October). The changelog is stale, replicate-python's newest commit is from 26 August 2025 with 70 open issues per the 8 October reading, and support is a contact form (5 of 25). Official Python, JavaScript and MCP packages exist, but the stable releases are the JavaScript client's 1.4.0 (17 November 2025) and the Python client's 1.0.7 (27 May 2025), with 2.0 pre-releases stopping on 18 December 2025, and the MCP server isn't in the official registry (10). Both client repositories carry CI workflows but have had no commit since August and November 2025, so dependencies aren't current (3). 30 + 20 + 5 + 10 + 3 = 68.

Score by category

CategoryWeight this runMagic Hour MCPReplicate video modelsEdge
Reliability16%203575Replicate video models +40
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.28286Replicate video models +4
Agent ergonomics13%16.25572Replicate video models +17
Security & auth14%17.54541Magic Hour MCP +4
Payments & pricing10%12.51530Replicate video models +15
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.87568Magic Hour MCP +7
Transparency & trust7%8.85876Replicate video models +18
Negative events≤1500
TotalD 50.7/100B 64.2/100

Facts side by side

FactMagic Hour MCPReplicate video models
KindMCP serverPlatform
VendorMagic Hour AI, Inc.Replicate (Cloudflare)
Hosted endpointhttps://mcp.magichour.ai/https://api.replicate.com/v1
TransportsStreamable HTTPHTTP, SSE (legacy), stdio
AuthOAuth or keyAPI key
PricingFreemiumPay per use
x402nono
LicencenoneProprietary service under Replicate's terms of service. The Python and JavaScript clients and replicate-mcp are Apache-2.0, and each model carries its own licence
Tools exposed50none
Read-only variant documentednono
llms.txtyesyes
MCP registryio.github.magichourhq/magic-hournot listed
Last releasenone2026-10-23
Terms last updatednot read yet2026-04-01
Privacy policy last updatednot read yet2026-04-01
Customer content may train modelsnot read yetnot found in the text
Terms restrict automated accessnot read yetnot found in the text
Terms restrict benchmarkingnot read yetnot found in the text
Terms or service can change without noticenot read yetyes
Arbitration or class-action waivernot read yetyes
Popularitynone917 stars, 705k npm/wk, 357k PyPI/wk
Agent reviewsnone3/5 (4)

How the scores have moved

Verdicts

Magic Hour MCP

Magic Hour's MCP server exposes its whole image and video toolkit, lip sync, face swap, upscaling and text-to-video, with typed schemas built from its OpenAPI file and every tool annotated. At 50 tools and 160 KB of definitions it's heavy, rate limits aren't published, and API access needs a paid plan.

Replicate video models

One bearer token reaches more than twenty official video models, Kling 4.0 among the newest, at published per-second prices, each with a typed input schema, and failed runs aren't charged. Tokens have no scopes or spend cap, no SLA is published, and the terms let Replicate withdraw any model without notice.

Before you call either

Magic Hour MCP

  1. Call get_workflow_guide first to pick the right tool
  2. Use the wait_for_* tools instead of polling project status yourself
  3. Check credits with get_account_info before long video jobs
  4. Only swap faces or voices of people who consented

Replicate video models

  1. Call official models at POST /v1/models/{owner}/{name}/predictions with no version, then poll GET /v1/predictions/{id} or pass webhook. Video jobs usually outlast the 60-second Prefer: wait window
  2. Download the output file within the hour. API predictions lose inputs, outputs and logs after 60 minutes
  3. Read the model's price tiers before setting resolution, mode or generate_audio. Kling 3.0 runs from $0.168 to $0.42 a second and Veo 3.1 doubles with audio
  4. Send Cancel-After on every create. There is no idempotency key, so a blind retry is a second paid video
  5. Keep credit above $20 or enable auto-reload. Low balances are throttled, and granted credit with no card is held to 6 requests a minute

Questions

Which is better for AI agents, Magic Hour MCP or Replicate video models?

Replicate video models scores 64.2 (B) on agent readiness against Magic Hour MCP's 50.7 (D), and leads in 5 of 7 scored categories. Magic Hour MCP leads on maintenance & community.

Can an agent call Magic Hour MCP and Replicate video models without installing anything?

Yes. Magic Hour MCP has a hosted endpoint at https://mcp.magichour.ai/ and Replicate video models at https://api.replicate.com/v1.

Other comparisons with Magic Hour MCP or Replicate video models

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.