Helixor Code gives Claude Code, Codex, Cursor, Antigravity, Gemini, and Grok a persistent, living structural code graph and stateful TODO memory over the Model Context Protocol (MCP) — running locally with zero cloud keys required to start.
Deep dive into our 6-part technical documentation: CLI Reference, Code Indexing, Local Desk Web UI, Menu Bar Tray, and Agent Rules.
Go to code.helixor.dev to create an organization and log in. Once registered, you can manage your team and access shared workspaces. You will connect this account to your local development environment during helixor init after installing the CLI.
Helixor Code operates across two flexible tiers designed to meet teams wherever they are in their AI adoption journey:
Augment Mode · Free · No Key
Attaches to Claude Code, Codex, Cursor, Antigravity, Gemini, or Grok via local MCP. The agent brings its model; Helixor provides structural code search (code.search, code.impact.analyze) and stateful TODO stacks (brain.todo). Saves 6×+ token overhead by substituting AST motifs for whole-repo file dumps.
Orchestrate Mode · Enterprise · BYOK
Runs checkpointed, resumable operational pipelines where specialized models handle specific steps (Claude or Gemini verifies plans, human approves at the gate, Codex or Grok implements and audits) on your own keys with audit trails and shared team memory.
See detailed architecture in How It Works or review the Documentation Overview.
Helixor Code runs on macOS (companion daemon + menu-bar tray app) and Linux (headless systemd service + HTTP status API). Requires Python 3.11+.
Standard Installation via PyPI (Recommended with pipx):
Air-Gapped / Offline Installation:
The installer automatically provisions an isolated environment at ~/.helixor-cli and links the binary onto your PATH. For full installation options and requirements, see the CLI Reference Docs.
helixor init)Run the guided, interactive setup command to configure your local companion, background daemon, and project roots in one step:
What helixor init configures:
--skip-login.~/code).LaunchAgent on macOS, systemd on Linux).Common Setup Flags:
Detailed flags and environment variables are documented in CLI Lifecycle Documentation.
Verify that your companion daemon and agent MCP registrations are fully operational:
helixor doctor validates CLI release train, background daemon status, source roots, MCP tool responsiveness, and cloud portal sessions. For quick runtime endpoint status:
Read more in Health & Diagnostic Commands.
Helixor Code integrates with any MCP-compliant agent, IDE, or CLI. Each agent receives specialized rule files and instruction sets:
See instructions for each agent runtime in Agent Rules & MCP Integration Guide.
Bring living structural indexes, code quality findings, and TODO stacks directly into your editor sidebar:
.vsix..zip.Helixor uses lazy on-demand indexing by default, but you can build or rebuild indexes explicitly:
helixor up automatically starts a background debounced filesystem watcher. Modifying files refreshes indexes incrementally.
Learn how AST indices, freshness hashes, and project roots work in Code Indexing Documentation.
Control the companion daemon lifecycle with simple commands:
macOS Menu-Bar App Features:
Explore the browser desk interface at Local Desk Guide and appliance status at Menu Bar Tray Guide.
Once connected, your agent has access to canonical Helixor MCP tools:
| Tool | Capability | Example Agent Prompt |
|---|---|---|
code.search |
Structural AST search (definitions, usages, bonds, motifs). | "Where is the JWT token verified in the auth middleware?" |
code.impact.analyze |
Impact analysis across APIs, models, tests, and call graphs. | "What components will break if I modify UserSession.role?" |
code.tests.find |
Reverse query: finds tests covering specific behaviors. | "Find test cases covering payment webhook retries." |
code.review |
Architectural quality and motif assessment. | "Review the auth package for anti-patterns and circular bonds." |
code.diff |
Structural concept and working-tree diffs. | "Show structural changes between main and this branch." |
| Tool | Capability | Example Agent Prompt |
|---|---|---|
brain.todo |
LIFO subject-scoped stack of TODO frames (push, peek, update, pop). | "Create a TODO stack for refactoring payments and mark step 1 done." |
brain.design |
Durable design documents with phase lifecycle tracking. | "Save our architectural design decisions for the database migration." |
workspace.get |
Reads shared context, pinned decisions, and channel updates. | "What decisions did the team pin in the release channel today?" |
workspace.act |
Posts updates and pins decisions for cross-agent team handoff. | "Post a summary of the auth refactor to the workspace channel." |
Complete schema definitions and parameter guides are listed in the MCP Tools Reference.
Collaborate across engineering teams on shared codebases with a spoke-and-hub architecture:
Read about enterprise network architecture in Hub-and-Spoke Overview.
One-Command Recovery (helixor repair):
If an agent integration goes stale, background processes conflict, or registrations mismatch, run:
helixor repair cleans up orphaned listeners, restarts the companion daemon, and re-validates MCP configurations across all installed agents.
Upgrading the CLI:
Clean Uninstallation:
See troubleshooting recipes in Reconciliation and Recovery Docs.
Rule of thumb
One-off question
Just query code.search. Structural lookup answers with exact AST bonds — no whole-repo token waste.
Nontrivial change
Plan with brain.design and push a brain.todo stack so engineering intent survives across context limits.
Governed SDLC
Run multi-LLM workflows with human approval gates, evidence certificates, and checkpointed rollbacks.
Enterprise Sales & Onboarding
Augment mode is free for individual developers. Talk to our solutions team to configure enterprise hub synchronization and governed multi-LLM SDLC pipelines.
