Updated July 2026: Higgsfield retired its Basic / Pro / Max plans, including the $9 entry tier, in early 2026. The current ladder is Starter $15/mo (200 credits), Plus $49/mo or $39 annually (1,000 credits), Ultra $129/mo or $99 annually (3,000 credits) and Business ~$89/seat. Higgsfield's pricing page renders client-side, so these figures are corroborated across five independent trackers rather than scraped — confirm on their site before paying.
Higgsfield does not publish a documented REST API with published per-call rates as of July 2026. What it does offer are three distinct access surfaces — the higgsfield.ai web app, an MCP server, and a CLI — plus Canvas and a "Supercomputer" mode, all metered by a shared credit system tied to your subscription tier. If you're trying to "call the Higgsfield API" the way you'd call OpenAI or Replicate, the honest answer is: that's not how this product is built, and understanding the actual surfaces will save you a wasted afternoon.
What "Higgsfield API" Actually Means Today
Searches for "higgsfield api" mostly reflect intent, not an existing product. There's no public docs page listing REST endpoints, authentication flows, or per-request pricing for Higgsfield. That doesn't mean you can't integrate Higgsfield into a workflow — it means integration happens through MCP and CLI, not through a conventional HTTP API you'd hit with curl and an API key from a developer console.
This matters for planning. If your team's roadmap assumes "call Higgsfield programmatically like any other AI API," you need to re-scope around what's actually documented: an MCP server for agent-to-agent tool use, a CLI for scripted/batch workflows, and the Canvas interface for compositing shots. All three still draw from the same credit pool as your plan.
The Access Surfaces, One by One
higgsfield.ai (web app)
The primary surface. This is where account management, model selection, and generation history live. It's also the only surface where Higgsfield's unlimited-access and free-generation features apply — more on that below.
MCP (Model Context Protocol)
Higgsfield exposes an MCP server, which lets AI agents and MCP-compatible clients (coding assistants, automation tools) invoke Higgsfield's generation capabilities as a tool. This is the closest thing to "API access" in the traditional sense, but it's scoped to MCP semantics — tool calls through an agent runtime — not a standalone REST contract you integrate directly into a backend service.
CLI
A command-line interface for scripting generations, useful for batch jobs or pipeline integration where you don't want a human clicking through the web app for every asset. Like MCP, it draws credits from your plan.
Canvas
A compositing/editing surface for arranging generated assets — closer to a creative workspace than a generation endpoint, but it consumes the same credit pool.
Supercomputer
Higgsfield's higher-throughput generation mode, positioned for heavier workloads. Documented as a distinct surface from the web app, also credit-metered.
The Critical Caveat: Unlimited Access Doesn't Travel
This is the single most important fact for anyone planning to build on MCP or CLI: Higgsfield's own pricing documentation states that unlimited models and free generations are accessible only via higgsfield.ai, not via MCP/CLI, Canvas, or Supercomputer.
In practice, that means any perk tied to "unlimited" on your plan stops at the web app's edge. If your integration plan routes generations through MCP or CLI — which is the natural choice for automation — you're spending metered credits on every call, full stop. Budget accordingly, and don't assume web-app economics carry over to a scripted pipeline.
Models You Can Reach Through These Surfaces
The model roster spans image and video generation, and it's genuinely broad — this is Higgsfield's strongest differentiator versus most single-model UGC tools:
| Category | Models |
|---|---|
| Image | Nano Banana Pro (incl. 4K), Nano Banana 2, FLUX.2 Pro, FLUX.2 Max |
| Video | Kling 3.0, Sora 2 Pro / Max (720p/1080p), Seedance Pro + Seedance Pro Fast (480p/720p/1080p), Higgsfield DoP Lite (720p), Soul 2.0, Cinema |
Which models are reachable from MCP/CLI versus only the web app isn't itemized in public docs beyond the unlimited-access caveat above — treat model-by-model surface availability as something to confirm in-product before you architect around a specific one.
Credit-Based Cost Model
Higgsfield prices in three consumer/prosumer tiers plus custom Business plans, all denominated in monthly credits with an in-tier slider to buy more headroom:
| Plan | Monthly Price | Annual Price | Credits |
|---|---|---|---|
| Plan | Monthly | Billed annually | Credits |
| Free | $0 | $0 | ~10/day |
| Starter | $15/mo | $15/mo | 200/mo |
| Plus | $49/mo | $39/mo | 1,000/mo |
| Ultra | $129/mo | $99/mo | 3,000/mo (scales to 9,000) |
| Business | ~$89/seat | ~$62/seat | 1,500/seat/mo |
(Source: higgsfield.ai/pricing, verified July 2026.)
Starter is credit-capped at 200 credits. Plus and Ultra both let you push past the base allotment via a slider, which is effectively pre-paying for more generation headroom without jumping a full tier. Business plans exist for teams needing more, but pricing isn't published — that's a sales conversation.
Estimating Real Cost for API-Style Workflows
Because there's no published per-call rate card, you can't back into a "$X per generation" number the way you can with a metered API. What you can do is reason from the tier math: if you're running MCP or CLI workflows daily, model your expected generation volume against the credit counts above, and remember that heavier models (video, higher resolutions) will burn through a credit pool faster than a single low-res image call. Start on Pro if you're prototyping an automated pipeline — Basic's fixed 200 credits will bottleneck fast once you're scripting instead of clicking.
If your use case is specifically turning creative briefs into finished, publish-ready ads rather than raw model access, it's worth comparing against tools built for that end-to-end job rather than a generation surface — see our AI UGC generator guide and AI video generator guide for how that category is priced and structured differently.
When to Use Higgsfield's Surfaces vs an End-to-End Ad Tool
Higgsfield's strength is model breadth and cinematic control — Kling, Sora, Seedance, FLUX.2, and more, all under one credit system, reachable via MCP/CLI for automation. That's the right call if your team needs to experiment across foundation models or build custom generation pipelines.
It's the wrong tool if what you actually want is "product link in, finished vertical ad out" with minimal setup. That's a narrower, different job, and tools built specifically for it (see /tools) trade model breadth for a much shorter path to a publishable asset. For a full breakdown of Higgsfield's plan structure and what each tier unlocks, see our Higgsfield pricing guide; for general pricing across the category, /pricing is a useful reference point.
Bottom Line
There's no public Higgsfield REST API with documented endpoints or per-call pricing as of this writing. What exists is a credit-metered ecosystem across the web app, MCP, CLI, Canvas, and Supercomputer — with the important restriction that unlimited/free access is a web-app-only perk. Plan your integration around MCP or CLI if you need automation, budget in credits rather than per-call dollars, and don't assume web-app pricing behavior extends to scripted workflows.