# Swallow vs Akur8 Deploy

> Swallow and Akur8 Deploy compared: where each rating engine fits, how rates reach production, testing and sign-off, how AI is used and when to pick which.

Published 9 January 2026. Category: Comparisons.
Source: https://www.swallow.app/post/swallow-vs-akur8-deploy

Akur8 Deploy makes most sense when your actuaries already build their rating plans in Akur8 and the goal is to get those plans live with the fewest handovers. Swallow makes most sense when your rates could come from anywhere (a rating guide, an Excel rater, a US rate filing or another modelling tool) and you want the whole product, not just the rates, in one tested model. They aren't mutually exclusive. Plenty of teams could model in Akur8 and implement, test and serve in Swallow.

## Akur8 Deploy in brief

Akur8 made its name with pricing modelling software that it describes as "powered by transparent machine learning", split into modules including Data, Risk, Demand and Rate. In October 2025 it added two products: Rate Repo, a versioned and documented home for rates, and Deploy, a cloud rating engine that takes Akur8 rates into production.

Akur8's own summary of Deploy is that teams can "import rate plans in minutes, deploy them in seconds, and quote in milliseconds". Rates are served through a real-time API that a policy administration system can call. Deploy supports testing before go-live, A/B tests, simulations, and routing strategies that split the client base into segments. Changes are versioned and logged, and changes that touch production can be reviewed and approved. In August 2026 Akur8 released an accelerator on the Guidewire Marketplace, so PolicyCenter customers can call Deploy for rating without redeploying PolicyCenter.

Akur8 says it has more than 350 customers in over 40 countries, with more than 3,000 actuaries on the platform every day. In October 2026 it launched Akur8 Agents, AI agents that support actuarial work inside the platform while actuaries keep the final decision.

## Swallow in brief

Swallow is an AI-powered pricing platform and rating engine for insurers and MGAs. Each model is a plain JSON file of steps that run in order: inputs, rate table lookups, calculations, transforms, exclusions, refers, excesses and endorsements. Every version moves from draft to approved to published, carries its own tests, and records who did what and when. The same engine prices quotes over a real-time API, through embedded forms and in voice and chat, and is built to scale to millions of quotes a day.

AI sits in the build. Describe the product and upload a spreadsheet and Swallow's AI sets up a first model, or start from a US rate filing. Claude, ChatGPT or Cursor can connect to the [Swallow MCP server](https://www.swallow.app/developer/mcp) to learn the format, draft and validate a model and run its tests. The MCP server can't create, edit or publish a project, so putting a model live always happens in the dashboard, by a person.

## At a glance

| | Swallow | Akur8 Deploy |
|---|---|---|
| **What it is** | AI-powered pricing platform and rating engine | Cloud rating engine inside the Akur8 pricing suite |
| **Where rates come from** | Rating guides, Excel raters, US rate filings or any modelling tool | Rating plans built in Akur8 |
| **Model format** | Readable JSON of ordered steps, with tests on each version | Not published in detail |
| **Product rules** | Exclusions, refers, excesses and endorsements sit in the same model as the rates | Public material centres on rates, simulations and routing |
| **Testing** | Tests on every version, expected against actual, CSV upload and AI-generated cases | Pre go-live testing, simulations and A/B tests |
| **Sign-off and audit** | Draft, approved, published; approvals by role, logged; every quote records its version | Versioned and logged changes, with review and approval |
| **AI** | AI first drafts, AI-generated tests, MCP server for Claude, ChatGPT and Cursor | Akur8 Agents across the platform; machine learning in modelling |
| **Channels** | Real-time REST API, embedded forms, voice and chat | Real-time API; Guidewire PolicyCenter accelerator |
| **Commercial model** | Talk to us | Not published |

## Where they differ

### Building the rates versus running them

Akur8 is at its best designing the rating structure from data. Its GLM and machine learning tools are well used by pricing actuaries, and Deploy shortens the trip from that work to a live price. Swallow doesn't fit models to claims data. It starts from a rating structure someone has already designed, wherever it came from, and turns it into a versioned, tested model that prices live quotes.

### How much of the product lives in the model

Deploy is organised around rating plans from Akur8. A product that's actually on sale needs more than that: decline and refer rules, excesses, fees, taxes, rounding and the wording that goes on the policy. Swallow keeps all of it in one model as ordered steps. Endorsement steps pick the forms and clauses for each quote, so the [policy documents](https://www.swallow.app/developer/policies) come from the same rules as the price.

### Proving a change is safe

Both products version their changes and let someone review them before release. In Swallow the proof is a test suite on every version. Each test is a full set of inputs with the result you expect, and the dashboard puts expected and actual side by side, so a rate change shows exactly which quotes moved and by how much. Every quote records the version that priced it, and a [debug trace](https://www.swallow.app/developer/quoting) lists every input, lookup, formula and rule along the way. Our post on [testing pricing models](https://www.swallow.app/post/testing-insurance-pricing-models) covers building a suite.

### Where AI does the work

Akur8 uses machine learning in modelling and now has agents that help actuaries inside its platform. Swallow points AI at the production model: a first draft from a description and a spreadsheet, tests generated from a prompt, and governance reports and version summaries for review. Over MCP, general AI clients can draft and check models against the same schema and tests your team uses, then hand the file to a person to upload, approve and publish.

### Cost

Akur8 doesn't publish its prices. For Swallow pricing, [talk to us](https://www.swallow.app/contact).

## Pick Akur8 Deploy if

- Your pricing team models in Akur8 already and wants rates to go from model to production without leaving the suite.
- You want A/B testing and routing strategies built into live rating.
- You're on Guidewire PolicyCenter and want a vendor-built accelerator for external rating.
- Your policy system already holds the product rules and you only need an external engine for rates.

## Pick Swallow if

- Your rates sit in a rating guide, a spreadsheet or a filing rather than in a modelling tool.
- You want rates and product rules together in one model that people can read and test.
- You want Claude or ChatGPT helping to build models, with a person approving and publishing every version.
- You need the same model quoting over an API, in embedded forms and in voice or chat.

If you don't have the capacity to build the model yourself, send us your rating guide, spreadsheet or filing and we'll talk you through getting it live.

## Running both

Many teams will keep Akur8 for modelling, and that works well with Swallow. Factors and relativities from Akur8 can be uploaded to a Swallow model as CSV rate tables, with the product rules built around them and tests run on every version.

## Next steps

Walk through a complete model on the [developer page](https://www.swallow.app/developer), see how quotes are priced through the [API](https://www.swallow.app/developer/api), or [get in touch](https://www.swallow.app/contact). For the wider market, read our comparison of [Swallow and other insurance pricing platforms](https://www.swallow.app/post/swallow-vs-insurance-pricing-platforms).

*Based on publicly available information as of November 2026. Vendors change their offerings, so confirm current details with each one.*