Fundraising Fox

RecurseML

Entrepreneur First '23

London, GB · Founded 2023 · 4 known investors

Recurse ML is an AI-powered code review tool that analyzes a team's codebase to automatically identify bugs, detect breaking changes, and suggest fixes before code is merged. It targets software development teams looking to reduce code review time, and links suggested fixes to relevant library documentation.

Also known as Cerebral Adaptive Forecasting Ltd. · Recurse · Recurse ML

Founders & leadership

RecurseML was founded in 2023 by Jack Jackson and Armins Stepanjans.

JJ
Jack JacksoninFounder2023–2026
ASArmins Stepanjans
Armins StepanjansinCo-Founder

Investors · 4

Company profile

researched Aug 2026

RecurseML (trading name of Cerebral Adaptive Forecasting Ltd.) develops an AI code review product that scans code changes for bugs before they are merged. The core workflow is a GitHub App: when a pull request is opened, Recurse analyses the code changes and leaves comments on the specific lines it believes are buggy. The company describes the service as language-agnostic across major programming languages and zero-configuration, working with default settings on the repositories to which it is granted access.

The product is positioned around three capabilities described on the company website: code analysis, "PR intelligence", and customisation. It claims to detect breaking changes — such as a modified function signature whose call sites were not updated, or configuration changes with production impact — and to suggest fixes consistent with the existing codebase. A second capability is external and internal library awareness: Recurse claims up-to-date knowledge of third-party libraries (React, Vue, Angular are cited as examples) and links suggested fixes back to the relevant documentation. A third is whole-codebase understanding, intended to surface issues both upstream and downstream of a change. Beyond the PR bot, the product ships as a CLI tool ("rml") and integrates with the Claude Code and Cursor coding assistants.

On data handling, the website states that Recurse does not train models on customer code, applies a zero-retention policy to analysed code, and is "currently pursuing" SOC 2 compliance. The GitHub App listing points to a privacy policy, terms of service, a list of data subprocessors and a data processing addendum.

Business model

Recurse sells a self-serve subscription developer tool. The site advertises a single Pro Plan at $25 with monthly or yearly billing (yearly billed as "2 months free"), preceded by a 14-day free trial. The plan bundles unlimited access to the "rml" CLI tool, integration with Claude Code and Cursor, and a GitHub pull-request review bot. The website's FAQ references a question about whether Recurse has an open source plan, but the answer text is not captured in the sources.

Subscription software priced at $25 per plan period (monthly or yearly, with yearly discounted by the equivalent of two months), following a 14-day free trial.

Traction

Company-reported figures on the website state 15,000+ bugs resolved and 80% less time spent debugging. The site displays a "trusted by" section and named testimonials from users at Continue.dev, Requesty, SurfSense, Robyn, Vizval, Capture.Energy and Findlay Park, including specific praise for SQL query analysis and catching edge cases. No user counts, revenue or funding figures are disclosed in the sources.

Latest developments

The company website (copyright 2026) foregrounds integration of Recurse with coding assistants Claude Code and Cursor, and states that SOC 2 compliance is being pursued.

Full profile — market position, technology, go-to-market, history, risks & controversies

Market position

Recurse operates in the AI code review tooling category and names CodeRabbit as a comparison point in its own FAQ. A third-party London startup directory lists RecurseML at seed stage in the AI category, though its description of the company differs from the company's own materials. No independent market share, revenue or ranking data appears in the sources.

Claimed differentiators, all from company-controlled sources, are full-codebase context (finding issues upstream and downstream of a change rather than only within the diff), current knowledge of external and internal libraries with links to the relevant documentation for each suggested fix, detection of breaking changes across call sites and configuration, language-agnostic zero-config operation, and a zero-retention/no-training-on-customer-code policy. The website FAQ explicitly includes a "How is Recurse different from CodeRabbit?" entry, indicating direct positioning against that competitor, though the answer text is not available in the sources.

Technology

An AI system that analyses pull request diffs in the context of the wider codebase to flag likely bugs, breaking changes and library misuse, then posts line-level comments with suggested fixes and documentation links. It is delivered as a GitHub App, a CLI binary ("rml"), and integrations with Claude Code and Cursor. The company states it is language-agnostic across major programming languages, requires no configuration, does not train on customer code, and retains no analysed code.

Go-to-market

Bottom-up, self-serve developer distribution: installation through the GitHub Apps marketplace (grant repository permissions and the bot begins reviewing PRs automatically), a CLI, and integrations with the Claude Code and Cursor coding assistants. Marketing relies on a 14-day free trial and named testimonials from founders and engineers at other technology companies, plus Twitter, Discord and GitHub community channels linked from the site.

Software engineering teams and individual developers that use GitHub pull requests and AI coding assistants. Testimonials on the website come from founders and engineers at startups and technology companies including Continue.dev, Requesty, SurfSense, Robyn, Vizval and Capture.Energy, plus an AI engineer at investment firm Findlay Park.

History

Sources indicate the product is distributed as a GitHub App under the developer account "Recurse-ML" and via a website operating under the legal entity Cerebral Adaptive Forecasting Ltd., trading as Recurse ML, with a copyright notice of 2026. No dated funding, acquisition or leadership events appear in the available public sources.

Risks & controversies

SOC 2 compliance is described as in progress rather than achieved, which may limit adoption by security-sensitive enterprise buyers. Nearly all available evidence — usage statistics, customer testimonials and product claims — originates from company-controlled channels and is not independently verified. A third-party directory entry describing RecurseML as running unit tests, performance benchmarks and bias checks on ML code in CI/CD conflicts with the company's own description of its product, indicating unreliable secondary listings.

Compiled by commissioned research from 6 cited public sources — announcements, filings, and press listed under research sources below.

Key figures

latest reported
Bugs resolvedJan 202615,000 bugs
Pro plan priceJan 2026$25
Reduction in code review / debugging time (company-reported)Jan 202680%

Company-reported or press-reported figures, each dated to when it was claimed — not independently audited.

Competitors · 5

by search overlap

Companies competing with RecurseML for the same Google search keywords, organic and paid, via search-intersection analysis.

In the news

Research sources · 6

primary sources listed

6 public sources were cited for this profile; the first-party ones are listed here.

Frequently asked questions

What does RecurseML do?
AI code review tool that scans pull requests for bugs and breaking changes and suggests fixes before code is merged.
Who founded RecurseML?
RecurseML was founded by Jack Jackson, Armins Stepanjans in 2023.
Who are RecurseML's investors?
RecurseML's investors include Entrepreneur First, Seedcamp, Twin Path Ventures, Playfair Capital.
Where is RecurseML headquartered?
RecurseML is headquartered in London, GB.