Influxx Supported CLIs: Claude Code, Codex, Grok, and the Catalog
Influxx Supported CLIs: Claude Code, Codex, Grok, and the Full Catalog — this guide is the deep product map of which coding agents Influxx treats as first-class citizens, how Claude Code, Codex, and Grok fit day-to-day fleets, how the wider catalog (Cursor Agent, GitHub Copilot, OpenCode, Gemini, Pi, Amp, Goose, Droid, Cline, Kimi, Hermes, Devin, and more) shows up in pickers, and how to set defaults, launch command overrides, and provider accounts without turning the cockpit into a second proprietary chat bot. If it runs in a terminal, it runs in Influxx — this article focuses on the actual CLI list and how to choose among them.
What “Supported” Means in Influxx
An Influxx agent is a terminal CLI (TUI) running inside a worktree-scoped PTY. Support means the app knows how to list, launch, title, track status, and surface usage for that binary — not that Influxx replaces the CLI’s own tools, models, or auth. First-class catalog entries get polished picker names and defaults. Any shell binary you configure still fits the same tab, isolation, and notification model. Prefer the local CLI path for production work: your keys, your subscriptions, your checkout.
This article intentionally goes deeper on the catalog itself. Launch entry points, parallel worktrees, note Run context packs, native chat chrome, and orchestration skills are covered in the agents and parallel runs help topic. Here you will leave knowing which CLIs exist in the product catalog, how Claude / Codex / Grok typically specialize on a team, how to set defaultTuiAgent and per-agent command overrides, and how provider accounts and trust presets keep multi-CLI days sane.
Bring the CLIs you already pay for — Claude Code, Codex, Grok, and the rest of the catalog — and let Influxx host them in real worktree terminals.
Set the Default Agent Before You Specialize
During first launch, pick any CLI agent Influxx knows as your default. Change it later under Settings → Agents (Coding agents). The default drives quick new-agent tabs and influences Sidebar note Run mapping when you do not pick another CLI. Sidebar Run commonly maps to the Claude TUI agent unless the binary resolves as another known TUI agent — for multi-CLI fleets, pick Codex, Grok, Cursor Agent, or others explicitly from New Workspace, a new agent tab, or the Influxx CLI.
Headline Trio: Claude Code, Codex, and Grok
Many teams standardize on a small primary set and keep the rest of the catalog for experiments. Claude Code is a common default for implementation work and note-driven runs. Codex (OpenAI Codex CLI) is a frequent peer for competing implementations and for workflows that benefit from Codex-shaped resume and history patterns when native chat is enabled. Grok (xAI) is the natural primary when your org standardizes on Grok for coding sessions or when you want a Grok-shaped opinion beside Claude or Codex on the same prompt across isolated worktrees.
- Claude Code: Strong default for plan-note Run, refactors, and general implementation TUIs; often the binary teams install first
- Claude Agent Teams: Team-oriented Claude workflows when you want multi-agent Claude patterns inside the same cockpit
- Codex: OpenAI Codex CLI for Codex-native stacks; pairs well with native chat resume/history where supported
- Grok: xAI Grok agent for Grok-standardized teams and multi-provider bakeoffs
Treat the trio as complementary, not exclusive. A typical week might run Claude Code from Sidebar notes for the main feature branch, spin Codex on a sibling worktree with the same prompt for a competing design, and reserve Grok for exploratory spikes or provider-diverse review. Influxx’s isolation unit is the worktree, so each CLI gets its own checkout when you want clean comparison — or share a checkout with a second terminal when collaboration on one branch is intentional.
Full First-Class Catalog (Non-Exhaustive)
Beyond Claude, Codex, and Grok, Influxx ships catalog entries that keep pickers honest for mixed fleets. The list is non-exhaustive and grows with the product, but shipped notes call out at least the following first-class names.
- Cursor Agent: Cursor’s agent CLI for Cursor-shaped runs inside Influxx worktrees
- GitHub Copilot: Copilot CLI entry for teams already in the GitHub agent ecosystem
- OpenCode: OpenCode catalog entry for OpenCode-centered workflows
- Gemini: Gemini agent entry when Google’s stack is part of your provider mix
- Pi: Pi catalog entry for Pi-based agent sessions
- Amp: Amp entry for Amp CLI users
- Goose: Goose entry for Goose-driven agent loops
- Droid: Droid catalog entry for Droid CLI sessions
- Cline: Cline entry for Cline-based coding agents
- Kimi: Kimi entry for Kimi agent workflows
- Hermes: Hermes entry for Hermes CLI runs
- Devin: Devin-related catalog entry where wired for launch from the cockpit
- Any shell binary: Custom wrappers, version-managed shims, and experimental agents via command configuration
Enable, Hide, and Override Launch Commands
Open Settings → Coding agents. Set the default agent. Enable or hide detected agents so the picker stays short. When a binary is missing from PATH, use whatever enable/install affordance the product offers for that agent, or install the CLI yourself and refresh detection. Launch command overrides replace or extend the default start line per agent — the right home for version managers, wrappers, non-standard install paths, and flags you refuse to retype. Related TUI options (permission styles, awake behavior, generated tab titles, status hooks, cache timers) live beside launch configuration so runtime comfort stays next to the binary definition.
- Default agent: Convenience for new agent tabs and note Run when you do not override
- Per-agent overrides: Stable launch commands for Claude, Codex, Grok, and every other catalog row you use
- Windows agent runtime: On Windows hosts, choose Windows vs WSL detection/launch so PATH matches reality
- Notes pane default: Configure which CLI Run uses for Sidebar notes when you want note launches decoupled from the general default
Provider Accounts, Usage Meters, and Trust Presets
Multi-CLI days are multi-account days. Settings → AI Provider Accounts (Providers) lets you switch Claude, Codex, Grok, Gemini, OpenCode Go, MiniMax, and similar setups without a full re-login where the product supports account switching. Status bar usage meters and rate-limit awareness for major providers help you see when a fleet is burning quota. Trust presets can pre-mark some agent workspaces so the first paste is not eaten by trust menus — they do not bypass agent sandboxes. Pair meters with Appearance status bar toggles so you only display the indicators you watch.
Dual intelligence still matters for catalog choice. Local CLI is the production path: worktree PTY plus your CLI and keys. Managed credits are the optional onboarding-friendly lane with chat-style runtime; a Pro seat unlocks cockpit access for bring-your-own CLI, notes, and orchestration, while credits meter only the managed path. Prefer local Claude, Codex, Grok, and peers for serious shipping. Managed is not a reason to ignore the catalog — it is an alternate lane when you want credits instead of every provider login on day one.
Choosing Among Claude, Codex, and Grok for a Task
Start from constraints, not brand loyalty. If the plan note is already written and you want the shortest Run path, Claude Code as default is often enough. If you need a second opinion on the same acceptance criteria, create another worktree and launch Codex or Grok with the identical prompt. If your org’s subscription and policy center on one provider, set that CLI as default and keep the others enabled for bakeoffs only. If native chat resume/history patterns matter for your primary agent, confirm that agent’s chat parity notes (Codex is called out for resume/history patterns) and decide whether tabs open in chat by default.
- Single-provider teams: One default, overrides for flags, catalog pruned to what you install
- Bakeoff teams: Same prompt across Claude / Codex / Grok worktrees; merge the winner
- Mixed specialists: Claude for implementation, Codex for alternate design, Grok for spikes — board statuses keep them honest
- Custom wrappers: Shell binary configuration when your internal agent is not a catalog name
How the Catalog Shows Up Across Entry Points
New Workspace composer: pick the agent explicitly with project, prompt, and base branch. Tab strip: New agent uses the default; Quick Commands can store alternate launchers. Sidebar note Run: context pack plus CLI tab — default mapping leans Claude unless configured otherwise. Tasks page: start from GitHub, Linear, or Jira with the agent you choose. Influxx CLI: worktree create --agent … --prompt … or terminal create --command … for scripted fleets. Automations: scheduled prompt runs with the agent wired into the automation. The catalog is the same list everywhere; only the entry point changes.
Example CLI shapes from product docs use patterns like creating a worktree with --agent codex and a prompt, then attaching another terminal with a command such as claude on the active worktree. Prefer --json when scripting or when another agent drives Influxx. Binary names for the Influxx CLI itself may be influxx, influxx-ide on some Linux PATH installs, or influxx-dev for dev builds — that is the cockpit CLI, not the coding agent catalog. Coding agents remain claude, codex, grok, and whatever override strings you configured.
Keep the Picker Honest
Hide agents you do not install so New Workspace stays scannable. Put long flag sets into overrides instead of Quick Commands when the flags should always apply. Use Quick Commands for occasional alternate launch lines. Revisit AI Provider Accounts when rate limits hit mid-fleet. On SSH remotes, remember detection and PATH belong to the remote machine — install Claude, Codex, Grok, or peers on the host where the PTY runs, not only on your laptop. On Windows, align agent runtime (Windows vs WSL) with where those binaries actually live.
Native Chat Is Presentation, Not a Different Catalog
Optional native chat chrome sits over the same agent PTY. Settings can open agent tabs in chat by default. Chat aims at TUI parity for supported agents — for example Codex resume and history patterns — so you are choosing presentation, not switching to a different product agent. Claude, Codex, Grok, and the rest of the catalog remain the same binaries whether you watch a raw TUI or chat chrome. Status, usage, and worktree isolation do not change when you flip presentation.
The catalog is a roster, not a cage — default Claude for speed, invite Codex and Grok for competition, keep every other CLI one override away.
Troubleshooting Catalog and Launch Issues
If an agent is missing from the picker, confirm it is enabled in Coding agents and installed on the runtime PATH (local or SSH). If launch fails, test the override command in a plain terminal tab. If note Run always opens the wrong CLI, check Notes pane defaults and the general default agent. If usage meters look empty, confirm provider account setup for that CLI’s ecosystem. If WSL agents mis-detect, set Windows agent runtime explicitly. If a custom binary never appears as a named catalog row, configure it as a shell command and launch via New terminal, Quick Commands, or terminal create — first-class naming is convenience, not a hard gate.
What “Done” Looks Like for CLI Setup
You are done with catalog setup when your default agent matches how you start most work, Claude Code / Codex / Grok (and any other CLIs you rely on) are installed on the machines where PTYs run, launch overrides encode the flags you need, unused catalog rows are hidden, provider accounts and usage meters cover the fleets you actually run, and you know when to fan the same prompt across multiple CLIs in isolated worktrees versus stacking two terminals on one checkout. Related help covers parallel runs, note context packs, settings privacy, and the setup checklist for wiring notifications and the Influxx CLI itself.
- Default set: Coding agents → default TUI agent chosen deliberately
- Trio ready: Claude Code, Codex, and Grok install paths verified where you use them
- Catalog pruned: Unused entries hidden; custom binaries overridden or Quick Commanded
- Accounts metered: Provider switching and status bar usage understood
- Remote PATH: SSH hosts have the same CLIs you expect locally when agents run remote
Quick recap: Influxx’s first-class catalog includes Claude Code, Claude Agent Teams, Codex, Grok, Cursor Agent, GitHub Copilot, OpenCode, Gemini, Pi, Amp, Goose, Droid, Cline, Kimi, Hermes, Devin, and more, plus any shell binary you configure. Set defaults and overrides in Settings → Agents. Use Claude, Codex, and Grok as a practical primary trio for note runs and bakeoffs. Prefer local CLI for production; treat managed credits as optional. Keep picker, PATH, provider accounts, and trust presets aligned with where sessions actually run — and let parallel worktrees decide when multiple CLIs should compete.
