R_
Reterm
Launch→
Reterm · v0.1.0 · $RTM on Solana · pending

I'm Reterm

An autonomous AI terminal commanding 9 specialized sub-agents across frontier language models. Not a chatbot, not an assistant — a self-improving multi-agent system rendered in pure monochrome.

Launch console→Read docsDiscover more↓
Count
9
Specialized sub-agents
01
Models
9
Frontier LLMs routed
02
Runtime
24/7
Always-on orchestration
03
Surfaces
5
Native channels
04
Orchestrator online◇9 sub-agents◇$RTM on Solana · soon◇Always learning◇Multi-model routing◇Token-gated access◇Stake · earn · govern◇Revenue share◇Self-improving◇Orchestrator online◇9 sub-agents◇$RTM on Solana · soon◇Always learning◇Multi-model routing◇Token-gated access◇Stake · earn · govern◇Revenue share◇Self-improving◇
§ 01 — AboutThe system

An autonomous AI orchestrator. Not a chatbot.

Reterm is a self-improving multi-agent system that dispatches work across nine specialized sub-agents, learns from every task, and settles incentives through the $RTM token economy on Solana.

Always-on01

Runtime

Powered by Reterm orchestration gateway running on Apple Silicon. Low-latency dispatch across sub-agents with zero cold starts.

◤◥
◣◢
→ see more
9 Models02

Multi-Model

Routes across nine frontier LLMs — GPT, Claude, Gemini, Grok, Llama, Kimi — choosing the right brain for each task in real time.

▢▢▢
▢▢▢
▢▢▢
→ see more
5 Surfaces03

Channels

Reach Reterm through CLI, web console, Telegram, Discord, and a native API. One brain, every surface, identical state.

─┬─
 ├ 
 ╰─
→ see more
§ 02 — Capabilities

Six primitives. Composed at will.

Every task Reterm executes is assembled from these six building blocks — combined, sequenced, and parallelised autonomously by the orchestrator.

Reason01 / 06

Multi-step reasoning

Chain-of-thought planning across heterogeneous models with intermediate verification at each step.

stable
12-layer CoT
Route02 / 06

Adaptive routing

Selects the lowest-cost competent model per sub-task. Latency, quality, and price are all live signals.

stable
9 model pool
Memory03 / 06

Persistent context

Vector + symbolic memory survives session boundaries. Reterm remembers what mattered, forever.

stable
∞ sessions
Tools04 / 06

Native tool-use

Filesystem, shell, HTTP, Solana RPC calls, and your own functions — registered once, used everywhere.

stable
Extensible
Govern05 / 06

Policy & guardrails

Declarative guards on cost, scope, and data egress. Every action is auditable and reversible.

stable
100% auditable
Learn06 / 06

Self-improvement

Outcomes feed back into routing weights. The orchestrator gets sharper with every run.

stable
Continuous
6 / 6 primitives · all stableOrchestrator ready
§ 03 — Console

One terminal. Nine minds at work.

reterm — orchestrator · zsh◇ live trace
status● running
agents3 / 9
latency142 ms
cost$0.041
§ 04 — Roster

Nine sub-agents. Each with a sworn duty.

Each agent runs on a dedicated frontier model, chosen for its specialty. The orchestrator dispatches, composes, and arbitrates. You give intent — they deliver.

Auditor
security
Static & semantic code review
Claude Opus
Reasoner
cognition
Multi-step planning & inference
GPT-5
Verifier
trust
Cross-model consensus checks
Gemini 2.5 Pro
Researcher
retrieval
Web + corpus retrieval
Grok 4
Writer
output
Long-form synthesis & prose
Claude Sonnet
Coder
execution
Code generation & repair
GPT-5 Code
Analyst
analysis
Numeric & data work
Kimi K2.5
Quartermaster
ops
Cost routing & budgets
Llama 3.1 405B
Sentinel
safety
Policy enforcement & guardrails
Reterm-tuned
All agents online · orchestrator v0.1.0Live
§ 05 — Console

Chat with Reterm. Live dispatch.

Connect your Phantom wallet and talk to the orchestrator. Reterm dispatches your prompt to the right sub-agent and streams a response.

reterm — orchestrator · solana
disconnected
[reterm] orchestrator v0.1.0 · 9 sub-agents on standby · connect wallet to dispatch
phantom wallet required to dispatch prompts
Reterm orchestrator · responses routed via sub-agent dispatch · Solana network
§ 06 — Questions

Frequently asked. Directly answered.

Reterm is an autonomous AI terminal that decomposes goals into sub-tasks and routes them across nine specialized agents and frontier language models. You give it intent; it returns verified results.

A chatbot replies. Reterm acts — it plans, dispatches, executes tools, verifies outputs across multiple models, and learns from outcomes. Conversation is one of many surfaces, not the product.

The roster currently includes GPT-5, Claude Opus 4.5, Gemini 2.5 Pro, Grok 4, Kimi K2.5, Llama 3.1 405B, and Reterm-tuned in-house variants. The Quartermaster agent picks the right one per task.

A free public tier exists. $RTM — a Solana SPL token — unlocks higher-tier agents, longer context windows, priority routing, and governance rights. Operators are paid in $RTM for running nodes.

Solana mainnet listing follows the v1.0 orchestrator release. Subscribe via the console for the gated allocation window.