Sixteen questions across four categories: the product, setup and operations, security and data privacy, and commercial terms.
CATEGORY 01
The Product
Grep searches text patterns and returns line matches; vector RAG chunks files and matches embeddings on semantic similarity. Both force the agent to open and re-read whole files to understand code relationships. Helixor Code parses 12 languages with Tree-sitter into a typed structural graph of definitions, usages, control and data flow, and weighted bonds. The agent queries typed facts and receives a compact ~150 token answer rather than thousands of tokens of raw file dumps.
Standard coding agents lose their entire working context the moment the terminal window closes or a session hits its token limit. Helixor Code maintains durable, depth-aware TODO stacks and workspace channels over MCP backed by persistent stores (brain.todo, brain.design). You or your teammates can resume a multi-day refactoring task across Claude Code, Codex, or different machines without having to re-explain architectural decisions.
When files are edited in git, Helixor monitors your working directory and splices working diffs into queries before answering. The agent is explicitly stamped with freshness status (e.g., freshness: stale in 2 files; git diff spliced) rather than hallucinating answers based on an outdated index.
Helixor Code supports 12 programming languages out of the box: Python, JavaScript, TypeScript, Go, Rust, Java, Kotlin, C, C++, C#, Ruby, and PHP.
A checkpointed multi-step SDLC workflow (sdlc.safe_change_delivery.v1) that routes each delivery phase to the optimal model on your own provider keys (BYOK). Phases like triage, planning, human approval, code generation, testing, and reviews require mandatory artifacts before advancing, with up to 3 automatic retries per step.
CATEGORY 02
Setup & Operations
About five minutes for a single developer. You run pipx install helixor and helixor init. The CLI automatically registers your source roots, starts the local companion, and configures detected agents.
No. The free Individual tier requires no API key, no credit card, and no subscription. Your AI coding agent (Claude Code, Codex, Cursor) supplies its own model credentials.
The companion is a lightweight local process running on http://127.0.0.1:18733. It serves the Model Context Protocol (MCP) endpoint at /mcp/, the browser-based Local Desk at /desk, and local health probes. It manages memory and query parsing without sending data to third parties.
Yes. Passing --no-tray and --no-desk allows Helixor Code to run headlessly in containers, remote SSH developer boxes, and automated CI pipelines.
CATEGORY 03
Security & Data Privacy
No. Proprietary source code and structural index artifacts stay strictly on your local workstation or CI server. Helixor never uploads your repository files to a cloud database.
Only team engineering memory: TODO stacks, architecture design summaries, channel messages, and server-derived author identity.
No. The local client uses standard outbound HTTPS connections to synchronize with your team hub. No open inbound ports or complex VPN tunnels are required.
Yes. Helixor Enterprise supports isolated on-premise deployments with complete airgap compliance and local network (LAN) hubs with zero dependency on public cloud endpoints.
CATEGORY 04
Commercial Terms
Yes. A single developer running on their machine has full access to structural code indexing, durable TODO stacks, and agent integrations at no cost forever.
Teams receive a full 3 months free from the day their hub goes live. After 90 days, pricing is $20 per seat per month for active developers attached to the hub.
Helixor Code never adds token markups. In augment mode, your agent uses your existing model account. In enterprise pipeline mode, all model calls execute on your provider keys (BYOK) at your contracted volume rates.
Have a question not answered here?
Reach out directly to the founders or explore the detailed technical guides.