CRISOL AI operates NORA, an autonomous software engineering platform that migrates legacy enterprise systems by planning, writing, testing, and deploying production code within the customer's own perimeter. NORA generates characterization tests from legacy systems in languages like COBOL, Natural, and PL/SQL, runs differential comparisons against modern implementations, and only proposes merges when equivalence is proven. The platform includes a self-hosted model family called VETA-1 for high-volume transformations without external API dependencies.
- Artificial Intelligence
- AI Agents
- Developer Tools
- Enterprise Software
- Financial Technology
- Government Technology
- Software Only
Funding
Founders
Product
Problem
Legacy enterprise systems written in COBOL, RPG, Natural, ABAP, PL/SQL, and other older languages contain undocumented business rules and accumulated technical debt that make modernization risky and slow. Organizations struggle to migrate these core systems because there is little evidence that new implementations behave identically to the old ones, and traditional manual rewrite projects take years and carry significant operational risk.
Solution
CRISOL AI operates NORA, an autonomous software engineering platform that migrates legacy core systems within the customer's own infrastructure perimeter. NORA plans, writes, tests, and deploys production code directly inside the customer's existing codebase and toolchain, with no data leaving the network. The platform generates characterization tests from the legacy system, executes both old and new versions in parallel against the same traffic, and only proposes a merge when a differential equivalence gate passes. This verification approach transforms the migration conversation from trusting AI output to reviewing hard evidence of behavioral equivalence.
Target Audience
Primary customers are banks, insurers, public administrations, and other European enterprises running critical legacy core systems that need verifiable migration to modern architectures under strict regulatory oversight including DORA and NIS2 compliance.
Features
- Orchestrated agent runtime with distinct roles: planner, cartographer, coder, verifier, reviewer, and documenter, running in ephemeral execution environments that are destroyed after each task
- Proprietary code intelligence layer that builds an AST graph of the entire estate, extracts undocumented business rules, and tracks data lineage across repositories
- Differential equivalence verification that generates characterization tests from legacy behaviorbots, runs parallel execution between old and new implementations, and gates merges on a green diff
- Legacy language and platform connectors for COBOL, RPG, Natural, Adabas, ABAP, PL/SQL, Oracle Forms, z/OS, CICS, DB2, IBM i, Java, and .NET
- VETA-1, a self-hosted model family post-trained on verified legacy-to-modern code pairs, deployable fully air-gapped for high-volume transformations without external API dependencies
- Eight deployment surfaces: cloud, desktop, console and API, IDE extension, Teams and Slack bot, MCP server, plus agent tooling for PR review, mass migration, incident triage, and security scanning
- Run-time supports micro-VMs without network access, with deployment options ranging from SaaS in the EU region to customer VPC to fully offline closed networks
- Integration with GitHub, GitLab, Bitbucket, Azure DevOps, Jira, Linear, ServiceNow, CI/CD systems, Datadog, and mainframe connectors