Skip to content

Enterprise

CodeNavi is built for the codebases that cannot leave the building. This page is the map for a platform, security or procurement team evaluating it: where the data stays, what ships, what you can audit, and who you are contracting with. Every claim links to the page that substantiates it.

Data boundary

  • Everything runs on hardware you control. Indexing, search, reranking and the call graph run inside one container on your host; the embedder is in the image and never phones home. There is no CodeNavi cloud in the path — Core concepts → The local boundary.
  • No outbound connections by default. The opt-in ones — a cloud LLM for chat, the crawler, webhooks, the update check — are enumerated in Network → Outbound connections. An air-gapped install has none.
  • Search never sends code anywhere. Only chat with a cloud provider sends retrieved chunks off the box, and the UI says so at the provider picker — Providers.
  • The embedder is trained once, by us, never on your data — The Buric-E4B embedder.

Deployment shapes

Shape Page
One developer, one box (Linux, Windows 11 + Docker Desktop, macOS) Setup guide, Windows, macOS
Shared team server — the host keeps its own clone, the index follows the repo, every agent queries over the LAN Team server
Air-gapped — carried package, offline signature verification, no update channel Air-gapped setup
Docker-free hosts (Fedora / RHEL, rootless Podman, SELinux enforcing) Podman
CI — a PR review bot calling diff_impact for the blast radius of a change CI review bots
No GPU on the host — build the index on a one-shot cloud GPU, run queries on CPU Cloud reindex, GPU compatibility

Sizing and the ~45 GB / 8–16 GB RAM / GPU tier questions: Setup guide → Hardware tiers. What running it actually costs a team in hours: How much work is it to run?.

Security & compliance

  • One scanner across the product — 48 rules for secrets, prompt injection and PII, applied at indexing, at the retrieval gate, in the crawler and in the wiki — Security model, Security features.
  • Evidence you can hand to an auditor — file:line traceability, a hash-keyed scan ledger, self-hashing report footers, and an honest statement of what integrity does not prove — Evidence integrity & auditability.
  • Compliance gap analysis against OWASP ASVS core, OWASP Top 10, or the EU Cyber Resilience Act's Annex I, run by your agent with CodeNavi's evidence-linked reports — Compliance gap analysis. The CRA report is an input to your Annex VII technical file; CodeNavi is supplier tooling inside your boundary and none of this is a certification.
  • Government guidance — a row-by-row mapping to the NSA MCP security design considerations and the joint Five Eyes agentic-AI adoption guidance, gaps included — NSA & Five Eyes guidance.
  • Verify the posture yourself — the curl-level checks and the audit hardenings behind them — Verify the security posture.
  • Vulnerability disclosure — Security disclosure.

Supply chain

Every release ships a CycloneDX SBOM (/app/sbom.json in the image), full third-party attribution (Third-party licenses), and signed packages — the installer verifies checksums and signature before anything runs, offline included. Model provenance (Buric-E4B, the ms-marco reranker, trust_remote_code=False) is in Setup guide → Model provenance. Updates are delta packages you apply yourself, with rollback — Updating CodeNavi.

Licensing & support

  • Licensor: Cellpex LLC d/b/a CodeNavi, an Ohio limited liability company. The software ships under the CodeNavi commercial EULA; the Buric-E4B weights under their own model license. Both are in your package (EULA.md, MODEL-LICENSE.md).
  • Trial: 30 days, full product, issued from codenavi.com. Licensing tiers and agency / multi-seat terms: ask through the same form.
  • Support: [email protected]. Include your /health output — Troubleshooting lists the first moves.
  • Rented-GPU hosting (vast.ai, RunPod, Hyperstack) is supported under an elevated trust model governed by Addendum A of the license — Rented GPUs.