Valmera vs Reap: two video MCP servers built for opposite jobs
Reap and Valmera are the two products most often named in the same sentence as "video editing MCP server", and they are close to opposites in what they do with the footage. Reap takes one long recording and multiplies it — clips, animated captions in 100+ languages, dubbing in 80+. Valmera takes one recording and finishes it — cuts, captions, music, grade, zooms, object removal, one export from the original file. Reap details were checked against Reap's own pages on 5 August 2026; Valmera details reflect the live product in October 2026.
If what you need is many deliverables out of one recording — short clips for several platforms, captions in a language your audience actually reads, a dubbed version for another market, all scheduled out to social accounts — Reap is the better product and it is not close. Valmera has no dubbing, no translation, no unattended batch delivery, no publishing and no team seats. If what you need is one video that is genuinely finished — dead air cut, captions timed to the new edit, music ducked under speech, a grade, punch-ins that land on the right words, a watermark repainted out — Valmera is the better product, and Reap does not attempt most of that. The MCP tool counts, 10 against 143, are not a scoreboard; they are the shape of those two bets written down.
Try Valmera on Your Footage
Upload real footage, describe the edit, get a finished video. Account creation and uploads are free; a subscription is required to edit.
Create an account →What each product actually is
Reap — a repurposing and localisation system
Reap describes itself as an AI video editor that turns long videos, webinars and interviews into clips, animated captions, dubbed versions and translated subtitles, and it is more accurate to read it as a production pipeline than as an editor. You give it a long recording; it finds the moments worth pulling, reframes them for vertical with speaker detection, captions them, optionally dubs them, and can publish them to TikTok, Instagram Reels, YouTube Shorts, LinkedIn, Facebook and X from a built-in content calendar. Its clipping is prompt-first — you direct it by intent, topic, hook or audience — rather than only algorithmic virality scoring, which matters if your reason for clipping is not virality at all.
The part that is genuinely uncommon is the language surface. Captions in 100+ languages, dubbing in 80+ with voice cloning, and explicit support for romanized scripts — Hinglish, romanized Urdu, Arabizi, romanized Bengali, Tagalog-English — which is how an enormous number of people actually write on social platforms and which most caption tools simply cannot produce. Reap reaches you four ways: the web app, a REST API, a CLI, and a hosted MCP server, plus connectors for n8n, Zapier and Make.
Valmera — an agent that performs the whole edit
Valmera is an agentic video editor. You upload real footage (up to 100 GB and 3 hours — MP4, MOV, M4V, MKV, WebM) and describe the edit in plain English. The system indexes the file into a word-level transcript, silence detection, shot detection and labeled frame tiles the agent reads directly, then the agent edits a versioned edit decision list, renders a preview, looks at the frames it just produced, and corrects itself. Nothing is destructive; any cut can be restored; the final export is cut from the original upload at source quality. The same tool registry the agent uses is published as a remote MCP server, so Claude, ChatGPT or Cursor can drive the whole thing from inside a conversation, through to the exported MP4.
Its scope is a deep edit of each project, including the shorts it cuts from a long video. That is the trade it makes: depth on each video instead of fan-out throughput across many.
Where Reap is genuinely better
This page is published by one of the two vendors, so the honest thing is to start here rather than bury it. Five areas where Reap wins outright:
Dubbing and translation. Dubbing across 80+ languages with voice cloning, and captions and subtitle translation across 100+. Valmera does neither, at all. If you publish the same video into more than one language, this single row decides the comparison and nothing further down changes it.
Romanized scripts. Hinglish, romanized Urdu, Arabizi. This is a real engineering choice, not a marketing line — producing captions in the script an audience writes in, rather than the one a language table says they should use, is the difference between captions that get read and captions that get scrolled past.
Volume. One request can produce many clips, each reframed and captioned, and Reap can publish and schedule them out. Valmera's Shorts workflow also turns one long video into multiple shorts, but each is reviewed and exported individually and there is no publishing integration. For a team whose job is filling a content calendar, the distribution step remains a meaningful difference.
Teams and brand consistency. Role-based workspace permissions, multiple seats, shared assets, and brand templates that lock fonts, colours, intros, outros and caption styles across everything the team ships. Valmera has no team seats, no collaboration and no stored brand kits — a repeated look has to be re-described each time.
A conventional API. Reap has a documented REST API with a published OpenAPI spec, plus a CLI, alongside its MCP server — and its own pages put API access on every paid tier from $9.99/month rather than behind an enterprise call. Valmera ships an MCP server and deliberately does not ship a REST API — auth is a user account over OAuth, there are no organisation seats and no per-project API keys to rotate. If you are wiring video into a backend service rather than into an agent, Reap fits that shape and Valmera does not.
And the entry price: Reap's paid tier starts at $9.99/month billed annually, below Valmera's entry plan (current Valmera plans).
Feature comparison
Reap rows were checked against Reap's own product, pricing, API and MCP pages on 5 August 2026. Where Reap does not publish something, the cell says not published rather than guessing — absence from a marketing page is not proof a feature is missing.
| Valmera | Reap | |
|---|---|---|
| Unit of output | One finished video, or shorts cut from it | Many clips / language variants |
| Starts from | Your uploaded file (100 GB / 3 h) | A long video or a URL |
| Remote MCP server | ✓ | ✓ |
| Tools exposed over MCP | 143 (live server card, October 2026), incl. 14 session tools | 10 |
| Final export over MCP | Yes — export_final | Yes — export_video |
| MCP auth | OAuth 2.1 + DCR, or bearer token | API key |
| REST API + CLI | No (MCP only, by design) | ✓ |
| Long video → short clips | ✓ | ✓ |
| Many deliverables per request | Multiple shorts from one video | ✓ |
| Publish / schedule to social accounts | ✗ | ✓ |
| AI dubbing (80+ languages) | ✗ | ✓ |
| Subtitle translation (100+ languages) | ✗ | ✓ |
| Romanized-script captions | ✗ | ✓ |
| Caption languages | Spoken language; English best-tested | 100+ |
| Caption styling | named presets · bundled fonts · entrances | 50+ presets |
| Subtitle file (SRT/VTT) export | ✗ | Translation published; file format not specified |
| Auto-reframe to vertical / square | 16:9, 9:16, 1:1, 4:5 | Yes, social formats |
| Silence and filler-word removal | ✓ | Not published |
| Colour grading and finishing effects | ✓ | Not published |
| Music library, ducking, mastering | ✓ | Not published |
| AI sound effects, images, short clips | ✓ | Not published |
| AI b-roll insertion | ✓ | ✓ |
| Erase a watermark or object from the pixels | ✓ | Not published |
| Agent can look at the frames it rendered | ✓ | No such tool published |
| Team seats and role permissions | ✗ | ✓ |
| Brand templates / stored brand kits | ✗ | ✓ |
| Free editing tier | ✗ | ✓ |
| Watermark on paid exports | None; paid exports end with a brief Valmera end card | None |
| Export quality | From the original file, source quality | 1080p (Creator) · 4K (Studio) |
Pricing
Reap's table was checked 5 August 2026 against Reap's own pricing page. Prices change; re-check before you buy.
Reap
| Plan | Price | What it includes |
|---|---|---|
| Free | $0 | 1 h/month of AI clipping and 1 h of captioning, 720p watermarked exports, 1 seat, no API access |
| Creator | $9.99/mo · $119.88/yr | Up to 10 h/month, 1080p watermark-free, up to 2 h/month dubbing and translation, 3 brand templates, 1 seat, API access |
| Studio | $29/mo · $348/yr | Up to 20 h/month, 4K watermark-free, up to 5 h/month dubbing and translation, up to 3 seats, 6 brand templates, API access |
| Enterprise | Custom | Custom hours, unlimited seats and brand templates, 4K, API access and custom integrations |
One thing worth knowing before you pick a tier: the monthly prices are shown as reductions from $24 and $50, so the headline $9.99 is a promotional rate rather than a list rate — treat it as such and check what it renews at. API access, on the other hand, is not gated the way it is at most tools in this category: Reap's own pages put the REST API, the CLI and the MCP server on every paid tier from $9.99, not behind Enterprise.
Valmera
Creating an account and uploading are free. Editing requires a subscription — see current plans. Edits use credits in proportion to the work they take, so media hours and credits are not directly convertible; compare the cost of completing a representative job, including revisions and delivery.
The useful comparison is not one sticker price against another. It is: what is the cost of the outcome you need? One finished, graded, music-mixed video a week is not something Reap produces at any price. Forty captioned clips in six languages is not something Valmera produces at any price.
The workflow difference, in one picture
The tool counts follow from the pipeline shapes. Reap runs a fan-out: one input, one analysis pass, many outputs, each finished by the pipeline and judged by you afterwards. Valmera runs a closed loop: one input, one edit decision list, a preview render, the agent looking at the frames it produced, and a revision — repeated until the single deliverable is right.
Read the two lanes as answers to one question: who checks the result? In a fan-out, the pipeline cannot check itself, because there is no single result to check — there are forty, and a human skims them. That is a perfectly sound design when the deliverables are short and interchangeable, and it is why Reap can be fast and cheap per output. In a closed loop, the agent is the first reviewer: it renders, it looks at the frames, and it fixes what it sees. That only pays for itself when the deliverable is one thing you care about, and it is why Valmera costs more per output and takes longer.
The tool counts fall straight out of this. Ten tools are enough to describe a fan-out — submit, poll, retrieve, plus one verb per output type. A closed loop needs a vocabulary rich enough for the corrections a reviewer makes: not "caption it" but "move that caption above the lower third for the six seconds the chart is on screen". That is where 129 editing tools come from, and it is the reason the number is not bragging — it is arithmetic.
The MCP surface, side by side
Reap claims to be the first video editing MCP server with integrated AI clipping. We have no evidence against that and no interest in disputing it — MCP is recent, Reap was early, and being early is worth something.
Its server is hosted at mcp.reap.video, documented as API-key based with an OAuth protected-resource discovery document published at the endpoint, installable in one command, and it lists ten tools: create_clips, add_captions, dub_video, translate_subtitles, transcribe, reframe, get_status, get_results, list_templates and export_video. Reap names Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Gemini CLI and ChatGPT as supported clients, and publishes an OpenAPI spec and an llms.txt so a model can read the docs. That is a well-built server. Six of those ten are editing verbs; the other four are the submit-poll-retrieve-and-deliver plumbing every async pipeline needs.
Valmera's server is hosted too, at https://valmera.io/mcp/server, and uses OAuth 2.1 with dynamic client registration and PKCE, so the client registers itself and you sign in — there is no key to store, paste into a chat window, or rotate — with a bearer token available for clients that have no OAuth support. It exposes 143 tools (live server card, October 2026): 129 editing tools in nine groups and 14 session tools for projects, uploads, indexing, rendering, review, export and download. An assistant can finish the job end to end: export_final, then wait_for_job, then download_url returns the full-quality MP4. Two design details matter more than the count. The registry is the same one Valmera's own agent uses — the tools are not re-declared for MCP, so there is no second list that can drift out of step with the product. And slow operations return a job id and a wait_for_job tool rather than a fabricated completion, because a model that is told a render finished when it has not will confidently describe frames that do not exist.
If you want the full field rather than these two, we keep a side-by-side of every video editing MCP server — Valmera, Reap, OpenClip, Video Jungle and the DaVinci Resolve bridge — and a definitional guide to what MCP video editing is. Setup for either server is a few minutes; ours is documented per client on the MCP setup page.
What this feels like on a real job
A 90-minute webinar you need to spread. Reap. You point it at the recording, it finds the moments, reframes them vertical, captions them, dubs the three best into Spanish and Hindi, and schedules them. You review a folder of finished clips. Doing this in Valmera would mean one conversation per clip and no dubbing at the end of it — the wrong tool, used stubbornly.
A 12-minute launch video that has to be right. Valmera. The dead air comes out and the cuts snap to word boundaries; captions are timed against the new edit rather than the raw recording; music ducks under speech; a grade goes on; punch-ins land on the words that carry the argument; the client logo in the corner of the screen recording gets repainted out of the pixels; the export is cut from the original file at source quality. Reap does not attempt most of that list, and would not claim to.
Both, in sequence. This is not a rhetorical concession — it is the most sensible pipeline for a lot of teams. Finish the long video in Valmera, then hand the finished file to Reap to clip, caption and localise. Both are remote MCP servers, so a single agent session can hold both connectors and move between them without you touching a file manager.
See what a closed loop produces
Upload one real recording and ask for the edit. Account creation and uploads are free; a subscription is required to edit.
Create an account →Where Valmera is genuinely better
Finishing. Cutting is the first ten minutes of an edit. The rest is captions timed to the new cut, music that gets out of the way of speech, a grade, zooms that land on the right words, a sound effect on the moment that needs one, and a watermark that has to go. Valmera has caption presets, fonts and animations, music found on the web with sidechain ducking, voiceover and one-request mastering to −14 LUFS, six colour-grade presets with custom exposure, contrast, saturation, temperature and tint, eight finishing effects, seven transition styles, and pixel-level erasure of a watermark or a small static object. Reap publishes none of that.
The agent sees its own output. After a render, Valmera's agent looks at the actual frames it produced and reacts to them. That is what makes "blur the username in the corner" or "end on the wide shot" land rather than approximate, because the aim point is a measurement off a real frame instead of a guess from a transcript.
Every change is reviewable. Each edit is a new project version with a rendered preview, so you can compare the result with the request, restore any cut, and approve the edit before you export.
Non-destructive by construction. Edits are entries in a versioned edit decision list, the upload is never modified, any cut can be restored, and the final export is cut from the original file — previews use a faster proxy, the deliverable does not.
Editing without a template. Because the tool surface is a vocabulary rather than a menu of pipelines, requests that no product has a button for still work: change the aspect ratio partway through and back, put text behind the subject, enhance a mouse cursor in a screen recording, splice in a generated still with Ken Burns motion, snap existing cuts onto the music's beat grid.
Where Valmera fits, and where it does not
Valmera is for one person with one video that matters, who would rather describe the outcome than operate a timeline. It is at its best on talking-head footage, screen recordings, podcasts and launch videos, in one language, ending in one file.
It is the wrong tool if you need dubbing, subtitle translation, SRT or VTT files, team seats, brand kits, direct publishing or scheduling, unattended batch delivery, a REST API, or a native mobile app — none of which exist. It is also not a text-to-video generator: it edits footage you already have. English is the best-tested transcription path. Captions are burned in. And a closed loop costs turns, which cost credits — if your measure of success is clips per dollar, Reap wins that measurement honestly.
Frequently Asked Questions
Choosing between them
Pick Reap if the video already exists and the problem is distribution: more clips, more platforms, more languages, on a schedule, with a team and a brand template holding it together. Its dubbing and its romanized-script captions are things Valmera cannot do at all, and its entry price is lower than ours.
Pick Valmera if the problem is the video itself: making one recording into something finished, with an agent that renders, looks at what it rendered, fixes it, and tells you the truth about what it changed. Account creation and uploads are free; editing requires a subscription.
And if the honest answer is both, run them in that order. Valmera makes the video worth spreading; Reap spreads it.
Made with Valmera: real exports
Start editing with Valmera
Upload real footage, describe the edit, review a finished cut. Account creation and uploads are free; a subscription is required to edit.
Create an account →