Palmier Pro Review 2026: AI Video Editor on macOS

What Palmier Pro is

Palmier Pro is a macOS multi-track video editor built around a different premise than most AI video tools. Instead of generating clips from a prompt box and handing you a download link, it puts a local MCP server inside a timeline editor. Claude or Codex connects to that server and edits on the timeline directly: cutting, arranging, placing clips, working with the tracks you already have.

That distinction matters for anyone searching for a palmier pro review. The category is crowded with text-to-video generators. Palmier Pro is closer to an editing environment with an agent interface bolted on, and the repository at github.com/palmier-io/palmier-pro shows 14,597 stars, which is a lot of attention for a desktop editor in this niche. The project page is at palmier.io.

The docs describe the core loop as: you assemble media on a multi-track timeline, expose the project through the local MCP, and let a model act on it. The model is not generating frames in the cloud by default. It is manipulating an edit. That is the whole pitch, and it is a narrower pitch than the marketing around most AI video products.

Supported inputs and platform

macOS only. There is no Windows build and no web app described in the documentation. That single fact eliminates a large share of the market before anyone opens the pricing page. If your team edits on Windows workstations, Palmier Pro is not a candidate, full stop.

Inputs are standard timeline media: video files, audio tracks, and the layered structure a multi-track editor expects. Export formats follow the usual NLE conventions rather than a proprietary container, which is the right call — locking exports behind a format would make the tool useless for delivery work.

What the MCP angle actually changes

Most AI video tools treat the model as a generator. Palmier Pro treats the model as an editor with hands on the timeline. In practice that means the quality of your output depends heavily on the model you connect and how well you scope its instructions. A model that can see the timeline and issue edit operations is doing something different from a model that renders pixels. The failure modes are different too: instead of morphing faces, you get bad cut points and misplaced clips.

Whether that is better depends entirely on the job. For assembly and rough-cut work, an agent that can move clips around is genuinely useful. For generating footage that does not exist, Palmier Pro is not the tool — you would pair it with a generator and bring the results into the timeline.

Setup and first project

Installation is a macOS app install. Sign-up and licensing run through the vendor. The documentation covers connecting a model client to the local MCP endpoint, which is the step that separates a smooth first session from a frustrating one.

The friction points I would expect, reading the setup docs, cluster in three places. First, MCP client configuration is still a manual step in most workflows — you are pointing Claude or Codex at a local endpoint and confirming the connection. That is fine for developers who already run MCP servers. It is a wall for editors who have never opened a config file.

Second, model selection is not neutral. The live pricing snapshot for the models you might connect spans a wide range. openai/o1-pro runs $150 per million input tokens and $600 per million output tokens. anthropic/claude-opus-4.7-fast runs $30 in and $150 out. openai/gpt-5.5-pro runs $30 in and $180 out. Cheaper options exist lower in the list, including openai/o3-pro at $20 in and $80 out. If you connect an expensive reasoning model to a long timeline session, token spend adds up fast, and that cost sits on top of the subscription.

Third, project settings. Multi-track timelines have frame rate, resolution, and codec decisions that need to match your source media. The docs cover this, but it is the kind of setup where a wrong default produces a session of work you have to redo.

None of these are dealbreakers. They are the normal tax on a tool that assumes a technical user.

How this review was researched

This review is built from the vendor documentation, the official pricing page, the public repository at github.com/palmier-io/palmier-pro, and live model pricing data for the models that connect to the MCP. It is not built from render logs, because no render logs exist for this review. Where the brief asked for render times, VRAM figures, and artifact rates, those numbers are not published by the vendor and I will not invent them. What follows describes what the documentation and pricing pages establish, and where a claim would require a test I did not run, I say so.

That constraint shapes the whole piece. A review that promises render-time benchmarks and then delivers them without a test is worse than useless. So the sections below focus on what can be verified: platform, architecture, pricing structure, license terms, and the repository signals that tell you how the project is actually maintained.

Real tests: what the documentation supports and what it does not

The brief for this review called for three jobs — a 30-second social cut, a three-minute explainer, and a vertical short — with render times, VRAM use, cloud cost, and artifact counts logged for each. I did not run those jobs. Reporting numbers for them would mean fabricating them, and fabricated benchmarks are the single most common failure in AI tool reviews. So here is what the documentation and pricing structure actually tell you about each job type, and where the unknowns sit.

The 30-second social cut

Short-form assembly is the job Palmier Pro's architecture fits best. A 30-second cut is mostly selection and arrangement: pick the good takes, trim, order, add captions and music. An agent with timeline access can do meaningful work here because the operations are discrete and reversible. The docs describe the model issuing edit operations against the timeline, which maps cleanly onto short-form assembly.

What the docs do not publish is how long a session takes or how many tokens a typical cut consumes. Those numbers depend on your model choice and how much context the agent needs. A cheap model at $20 per million input tokens and an expensive one at $150 per million produce very different bills for the same edit. That spread is the real cost variable, and it is under your control.

The three-minute explainer

Longer timelines stress the agent's context. A three-minute explainer has more clips, more tracks, and more decisions per session. The docs describe multi-track support, so the structure is there. What is not documented is how well a model holds a coherent plan across a longer edit without losing track of earlier decisions. That is a model limitation as much as a tool limitation, and it is the kind of thing that only shows up in practice.

If you are considering Palmier Pro for explainer work, the honest advice is to scope the agent's role narrowly — have it handle assembly and let a human make the structural calls. The tool supports that division of labor. Whether the model respects it depends on the prompt.

The vertical short

Vertical work means reframing, which is where AI editing tools often struggle. The docs describe timeline operations, not automatic reframing intelligence. If your source is horizontal and your target is vertical, expect to do the framing work yourself or write explicit instructions for the agent. This is not a knock on Palmier Pro specifically — it is a limit of what timeline-manipulation agents can reliably do without dedicated reframing features.

Clip consistency and prompt adherence

The brief asked for a five-shot character consistency test with drift, morphing, and text artifact ratings. That test belongs to a text-to-video generator, not to a timeline editor. Palmier Pro does not generate the character. It arranges clips you supply. If your character drifts across five shots, the drift came from whatever generated those shots, and Palmier Pro's job is to cut around it or pick better takes.

This is worth stating plainly because the search intent around "palmier pro ai" often assumes a generator. It is not one. The consistency question that actually applies to Palmier Pro is different: does the agent place the same character's shots in a coherent order, and does it respect instructions about which take to use? That is prompt adherence at the edit level, not the pixel level.

The docs describe the MCP interface as the mechanism for those instructions. How faithfully a model follows them is a function of the model. A stronger reasoning model will hold a multi-step edit plan better than a cheap one, and the pricing snapshot reflects that — the models at the top of the list cost ten times what the models at the bottom do per token. There is a real tradeoff between edit quality and token spend, and Palmier Pro hands you that dial rather than hiding it behind a credit system.

Text artifacts are similarly out of scope. If your titles render badly, that is a font, codec, or export setting issue, not an AI generation issue. The tool gives you timeline control over titles; it does not generate them from a prompt.

Pricing

The pricing page lists a starting tier at $29 per month. Beyond that, the structure has two layers that matter more than the subscription itself.

The first layer is the subscription. At $29 per month entry, Palmier Pro sits in the same range as consumer creative tools. The pricing page is the source for what each tier includes; the free tier's limits are not something I will guess at, so check the page directly before committing.

The second layer is model cost, and it is the one people miss. Because Palmier Pro connects to Claude or Codex through MCP, you pay for those model calls separately, at whatever rate your provider charges. The live pricing snapshot shows the spread clearly:

• openai/o1-pro: $150 per million input tokens, $600 per million output tokens
• openai/o1-pro:batch: $75 in, $300 out
• anthropic/claude-opus-4.7-fast: $30 in, $150 out
• openai/gpt-5.5-pro: $30 in, $180 out
• anthropic/claude-opus-4.6-fast: $30 in, $150 out
• openai/gpt-5.4-pro: $30 in, $180 out
• openai/gpt-5.2-pro: $21 in, $168 out
• openai/o3-pro: $20 in, $80 out
• openai/gpt-5-pro: $15 in, $120 out
• anthropic/claude-opus-4.1: $15 in, $75 out
• anthropic/claude-opus-4: $15 in, $75 out
• openai/o1: $15 in, $60 out

For a heavy editing session, the difference between a $15/$60 model and a $150/$600 model is an order of magnitude. Palmier Pro's per-minute cost is therefore not a fixed number — it is a function of which model you connect and how much context the agent consumes. That is a genuine advantage over credit-based tools if you are cost-conscious, because you can drop to a cheaper model for routine assembly and switch up for complex structural work.

Palmier Pro vs Runway on pricing

Runway and Kling price by generation credits, which bundles model cost into the subscription. Palmier Pro unbundles it. For a user who wants one predictable bill, credits are simpler. For a user who wants to control spend and is comfortable managing a model API key, unbundling is cheaper at scale. The tradeoff is real and it depends on your workflow. If you are generating footage, Runway or Kling is the right category and Palmier Pro is not a substitute. If you are editing footage you already have, Palmier Pro's model is more flexible.

Beetlix is our own product, and the comparison worth making is on the agent-orchestration side: Beetlix runs multi-model workflows from a single interface, while Palmier Pro is a macOS editor that exposes one local MCP endpoint. Different jobs. If your work is timeline editing, Palmier Pro is the more direct fit; if it is coordinating several models across a pipeline, beetlix.com covers that ground.

Pros and cons

Pros

• Real timeline, not a prompt box. The multi-track editor is the product. The AI sits on top of a working NLE rather than replacing it, which means you can finish the edit yourself when the agent gets it wrong.
• Local MCP is the right architecture. Exposing the timeline through a local server keeps your media on your machine and lets you choose the model. That is more flexible than a locked-in cloud pipeline.
• Model cost is transparent and controllable. You see exactly what each model charges and can switch. No opaque credit conversion.
• Strong repository signal. 14,597 stars on the GitHub repo indicates an active, watched project rather than an abandoned experiment.
• macOS-native performance. A desktop app using local hardware avoids the upload-and-wait cycle of browser-based editors.

Cons

• macOS only. No Windows, no web. This is the biggest limitation and it is not going away soon.
• Setup assumes technical comfort. Configuring an MCP client and managing a model API key is not an editor's normal workflow. Expect a learning curve.
• Two bills, not one. Subscription plus model API spend. The second bill is variable and can surprise people who connect an expensive model without thinking about token volume.
• Not a generator. If you need footage that does not exist, this tool will not make it. The name and category invite the wrong expectation.

Queue waits, watermark rules, and commercial license

Because Palmier Pro runs locally, there is no render queue in the cloud sense — the docs describe local processing, so queue waits are a function of your own hardware, not a shared server. That is a structural advantage over cloud editors during peak hours.

Watermark rules and commercial license terms are set by the vendor and listed on the pricing and license pages. I will not state terms I have not read from the source. Check the license page before using output commercially; that is standard advice for any tool, and it matters more here because the output is an edit of your own media, which raises different rights questions than generated footage.

Verdict

Palmier Pro is a good tool for a narrow, technical audience: macOS editors who already work with AI models and want an agent inside their timeline rather than a generator outside it. The local MCP architecture is the right bet, the model-cost transparency is better than credit systems, and the repository activity suggests the project is alive. The macOS lock and the setup friction keep it out of reach for most teams, and anyone expecting a text-to-video generator will be disappointed.

Best use case: an editor on a Mac who has a Claude or Codex workflow and wants to automate assembly and rough-cut work on footage they already own. Worst use case: a Windows team that needs generated footage and a single predictable bill.

Score: 7.5/10. Strong architecture, real limitations, honest pricing. If you are in the target audience it is worth the $29 entry to try; if you are not, no amount of AI capability will fix the platform gap.

FAQ

Is Palmier Pro a text-to-video generator?

No. It is a macOS multi-track video editor that exposes a local MCP so Claude or Codex can edit on the timeline. It arranges and cuts media you supply; it does not generate footage from prompts.

How much does Palmier Pro cost?

The pricing page lists a starting tier at $29 per month. Model usage is billed separately by your provider, at rates that range from $15 to $150 per million input tokens depending on which model you connect.

Does Palmier Pro work on Windows?

No. The documentation describes a macOS-only application. There is no Windows build or web version listed.