LLM-powered code migration and legacy-modernization startups

1. Grit — AI-powered code migrations and dependency upgrades that open PRs across large codebases.
Why: YC-backed, raised seed funding, and is explicitly built for automated legacy cleanup rather than general copilot features.
2. Codemod — AI-assisted codemod automation platform for framework migrations and large-scale codebase changes.
Why: Early product with a community of reusable codemods; directly targets the migration workflow that enterprises struggle with.
3. Second — AI agents that perform codebase migrations, refactors, and dependency upgrades and deliver pull requests.
Why: Very early, YC-backed startup focused squarely on “codebase maintenance as a service,” not just code generation.
4. Codegen — AI coding agent focused on large-scale legacy code migration and enterprise modernization.
Why: Started by ex-Meta engineers, has raised venture funding, and positions itself as an agent for complex refactors rather than autocomplete.
5. Mutable.ai — AI codebase assistant that automates refactors, documentation, and unit tests in existing repos.
Why: Early-stage and product-led; useful for modernization teams that need to understand and improve legacy code before migrating.
6. Moderne — Mass code refactoring and migrations using OpenRewrite recipes, increasingly with AI assistance.
Why: Established in the refactoring niche and has enterprise traction, but still relevant as an LLM-era modernization platform.
7. vFunction — AI-driven architectural discovery and transformation of legacy Java/.NET monoliths into microservices.
Why: Later-stage but directly addresses legacy modernization with real enterprise customers; strong signal of demand for AI-assisted architectural refactoring.
8. Mechanical Orchard — Uses AI and modern cloud engineering to reverse-engineer and rebuild legacy enterprise systems.
Why: Early but has attracted enterprise contracts; represents the services-to-product approach to modernizing old systems.
9. Sourcery — AI refactoring tool for Python that cleans up legacy code and reduces technical debt.
Why: Narrow but real product focused on automated code improvement, useful for modernization pipelines before larger migrations.
10. AppMap — Maps legacy application runtime and architecture, then provides AI-assisted refactoring recommendations.
Why: Open-source adoption and early traction; solves the “understand the legacy system first” problem that blocks most migrations.

White-space gaps I noticed:

• Correctness-guaranteed COBOL/mainframe to cloud rewrites. Most tools translate syntax or help engineers refactor, but few productize business-logic extraction, automated behavioral equivalence testing, and parallel cutover for mainframe workloads. There is room for a startup that sells outcome-based modernization, not just developer tooling.
• Productized modernization for non-tech enterprises. Lots of startups target engineering teams, but few offer a managed, fixed-fee or outcome-priced service for ERP customizations, VB6, Delphi, or niche legacy systems inside insurance, logistics, and manufacturing. The gap is packaging LLM migration into a compliance-aware, audit-ready modernization product for buyers without large internal platform teams.

Sourcing list — LLM-powered code migration / legacy modernization

1. Mechanical Orchard — Rewrites mainframe systems by observing runtime behavior and regenerating logic in modern stacks, increasingly LLM-assisted. Fits: Series B (Alphabet-backed), founded by Pivotal's Rob Mee; strong enterprise mainframe traction but still early in revenue scale.
2. Bloop — Pivoted from code search to automated COBOL-to-Java conversion using LLMs plus verification. Fits: seed/Series A-stage (YC, Paul Graham-backed), sharp wedge into banking/insurance mainframe estates.
3. Moderne — Automated large-scale code refactoring built on OpenRewrite, layering LLMs on deterministic AST transformations. Fits: Series A/B; the "recipes + AI" hybrid is arguably the most defensible architecture in the category.
4. Grit (grit.io) — Agentic system for automated code migrations and dependency upgrades (GritQL + LLMs). Fits: seed-stage YC company, early design partners doing framework migrations at scale; talent-dense small team.
5. Second (second.dev) — AI agents that execute framework and version migrations (e.g., Angular→React, major version upgrades) end-to-end. Fits: YC-backed, seed stage, founder previously built dev-tools; narrow migration wedge with usage-based pricing.
6. Codemod (codemod.com) — Platform and registry for AI-assisted codemods, community-driven migration campaigns for large codebases. Fits: seed stage, open-source distribution motion, partnerships with framework maintainers (e.g., Next.js ecosystem migrations).
7. Blitzy — Autonomous multi-agent platform generating and modernizing large enterprise codebases with heavy pre-compute/planning. Fits: early-stage, Boston-based, claims million-line codebase handling; enterprise pilots in modernization.
8. TurinTech (Artemis) — "Evolutionary AI" plus LLMs to optimize and modernize existing code for performance and cost. Fits: London-based Series A, fintech customer base, differentiated angle (optimization, not just translation).
9. CloudFrame — Automated COBOL-to-Java conversion tooling, now integrating GenAI for refactoring and documentation. Fits: lesser-known, bootstrapped-lean, real mainframe customers—interesting as an AI-native re-founding or partnership target.
10. Kodesage — Budapest-based startup using LLMs to document, comprehend, and modernize legacy enterprise systems (knowledge extraction before rewrite). Fits: pre-seed/seed, European enterprise focus, tackles the "no one understands this codebase" problem first.

White-space gaps:

1. Verification layer as a standalone product. Everyone generates translated code; almost no one sells independent equivalence-testing/behavioral-diffing infrastructure (runtime trace comparison, property-based test synthesis) that de-risks any vendor's migration. This is the trust bottleneck holding back nine-figure mainframe deals.
2. Data & schema migration alongside code. Nearly all players translate application logic but leave hierarchical databases (IMS, VSAM), stored procedures, and embedded business rules in data layers untouched — often 50%+ of modernization cost. An LLM-native "data-layer modernization" company is conspicuously absent.