How Terminal X routes one prompt across 35+ AI models

Terminal X is a multi-AI router. You write one prompt; the parser picks the right AI (or combination of AIs) for the job and runs them in parallel where it can.

Type a command in the box at the bottom. The brain view above shows which AIs are routed and lights them up as they execute. Results stream into the right-hand panel.

Try one of the example prompts on the welcome screen of a fresh session to see multi-AI orchestration in action.

Terminal X runs in your browser at terminalxapp.com and as an app on the iOS App Store. Your account, sessions, credits, and connections sync across both, so you can start a prompt on your phone and pick it up on the web.

Every command goes through a parser (Claude Sonnet 4.6) which returns a task plan as JSON. Tasks can be parallel or sequential — when one task depends on another (e.g. an HTML page that needs generated images), the parser sets a dependsOn field.

The executor then:

1. Runs tasks with no dependencies in parallel.
2. Waits for each child's parents to finish, joins their outputs into the child's context, and runs.
3. Streams every chunk back to the UI as it arrives via Server-Sent Events.

When Claude is the orchestrator (and other models depend on it), it switches into creative-director mode and writes labeled sections like [FLUX]...[/FLUX], [ELEVENLABS]...[/ELEVENLABS] — each section feeds the matching downstream model as its prompt.

Terminal X routes across 30+ AI models grouped by modality. Real API integrations live for ~half of them today; the rest fall back to a working alternative behind the scenes.

Plans give you a monthly credit allotment. Every routed model call costs a fixed number of credits. Basic $5 / 400 · Starter $19 / 2,500 · Pro $49 / 6,500 · Business $149 / 19,800. Every plan can use every model and every connector. There is no free plan — an account without a subscription has 0 credits.

Credits are deducted only for successful tasks. Failed calls cost you nothing. Unused credits roll over up to your plan's rollover cap (Basic 100, Starter 500, Pro 1,500, Business 5,000).

The pre-flight credit estimate shown before each command is a fair upper-bound — real cost is metered from actual token / character / duration usage and is almost always lower than the quote. You will never be charged more than the quoted amount.

See /billing for the full credit cost per model and plan comparison.

The mode picker next to the send button biases the router toward a tier of models:

• Default — best model for each task.
• Fastest — lowest-latency models (Haiku, Gemini Flash, Grok Fast).
• Cheapest — lowest-cost (DeepSeek, Llama, Qwen, Mistral, Flux).
• Best — highest quality (Opus, GPT-5, DALL-E, Runway, ElevenLabs HD).

Your selection persists in localStorage across sessions.

A Team is a named bundle of AIs you've curated for a specific workflow — e.g. a Marketing team with Claude + Flux + ElevenLabs + Suno. Manage teams at /teams.

When a session is bound to a team, the router strictly limits its picks to that team's AIs. If the user asks for something the team can't serve (e.g. a video when the team has no video models), the system surfaces a clear error: "Team 'X' has no AI for this. Add one or switch teams."

Switch the active team for a session via the team pill in the composer (next to the mode picker).

Terminal X extracts useful facts from your conversations and injects them as context for future commands. Memories are private and never used to train models.

You can also set custom instructions in Settings → Preferences. They get prepended to Claude's system prompt on every command, so things like "I'm a backend engineer, prefer TypeScript, keep answers concise" persistently shape every response.

Toggle memory on or off in Settings → Memory. When off, no new memories are extracted and existing ones are not used as context.

Every user has a unique referral code. Share your link (Settings → Referrals) and when an invitee upgrades to a paid plan for the first time:

• You get 500 credits
• They get 500 bonus credits on their first paid month

Rewards trigger automatically via the Stripe webhook. Self-referrals are blocked at signup.