Keep your data on your own infrastructure while leveraging our advanced code intelligence. Code passes through for processing but is never stored.
You run the collector and vector database. We handle the intelligent processing.
Runs on your servers
Managed by us
Your collector reads code from your repos
Code sent to our API, processed in memory
Only embeddings returned, code discarded
Vectors stored in your Qdrant instance
Monthly quotas for chunks processed and queries. Upgrade anytime as your codebase grows.
Small teams, single repo
Growing teams, multiple repos
Large teams, org-wide
Custom scale, custom needs
A chunk is a semantic unit of code—typically a function, class, or logical block. Our AST-based parser intelligently splits your code so each chunk maintains context and meaning. A typical 10,000 line codebase might have 500-1,000 chunks.
Python package you install on your infrastructure. Scan repos, send to API, store vectors locally. Fully auditable.
pip install repoengine-collector
Send code chunks, receive vectors. AST parsing, semantic chunking, embedding generation—all handled by our API.
Convert search queries to vectors for semantic search against your local Qdrant instance.
Track chunks processed, queries made, and quota usage. Real-time visibility into your consumption.
DPA, Terms of Service, Privacy Policy, and Security Whitepaper. Business tier and above.
Email support for Starter. Slack channel for Pro. Dedicated support for Business and Enterprise.
No. Code is processed in memory and immediately discarded after generating embeddings. We never write your source code to disk, database, or any persistent storage. Only the vector embeddings are returned to your collector, which stores them in your own Qdrant instance.
You need to run two things: (1) the RepoEngine Collector (a Python package) and (2) a Qdrant instance for vector storage. Both can run on a single server, in Kubernetes, or as managed services. Qdrant Cloud is a good option if you don't want to self-host.
As a rough estimate: 1,000 lines of code = ~50-100 chunks. A medium-sized monorepo (100K lines) might have 5,000-10,000 chunks. The Pro tier (50K chunks/month) handles most growing teams. Re-indexing counts as new chunks, so plan accordingly if you re-index frequently.
Yes. The collector is open source—you can inspect exactly what it sends. You can also enable debug logging to see every API request. For Business and Enterprise tiers, we provide detailed API audit logs.
With our cloud offering, we handle everything—indexing, storage, and the web interface. With hybrid, you run the collector and vector database yourself, and only use our API for processing. Hybrid gives you more control over your data but requires more setup.
No. Our indexing algorithms are proprietary and we don't distribute them. The hybrid model is our solution for organizations that need data residency—your code and vectors stay on your infrastructure while leveraging our processing capabilities via API.
Get started with a pilot or talk to our team about your specific requirements.