CHOOSE YOUR TIER
The local Merlin harness is open source and free forever. The Sanctum web workspace runs on three plans: start free with Jie, unlock Merlin and Nxelle Codex on every Flash model with Medium, or open the whole fleet on every AI model with High.
Free
Light, sharp Q&A with Jie — no card, no trial clock.
- Jie — short, sharp, substantive answers
- Engine: DeepSeek V4.1 Flash
- 30 messages / day
- Upload up to 2 text/code files (100 KB each)
- Voice input (mic dictation)
- Merlin — unlocks on Medium
- Nxelle Codex — unlocks on Medium
Medium
Merlin, Nxelle Codex and Jie on every Flash engine.
- Everything in Free
- Merlin — strategy, architecture & deep reasoning
- Nxelle Codex — dedicated coding engine
- All Flash models — DeepSeek V4 / V4.1 Flash, GLM 5.3 Flash / FlashX, Gemini 3.7 / 3.8 Flash
- 300 messages / day
- Upload up to 5 files (300 KB each)
- Frontier models (Opus, GPT, Gemini Pro) — High
High
Every agent on every AI model — frontier engines included.
- Everything in Medium
- Every AI model — Claude Opus 5, GPT 6 Astra / 5.6, Gemini 3.1 Pro, DeepSeek V4 Pro, GLM 5.3, Kimi K3, MiniMax M3 & more
- 1,000 messages / day, incl. 100 frontier-model messages
- Upload up to 10 files (1 MB each)
- Priority email support (48h)
- Early access to new engines
Enterprise
Dedicated sovereign deployment isolated in your own VPC.
- Self-Hosted Relay Container — our code in your air-gapped VPC
- Unlimited operators, machines, and agents
- RBAC, SSO & immutable audit log
- Custom DPA, security audits, and procurement paperwork
- Dedicated private Slack/Discord channel (4h SLA)
Jie
Free · Medium · High
Light reflex Q&A, short, sharp and substantive. Powered by GLM 5.3 Flash & Gemini Flash.
Merlin
Medium · High
The Governor: strategy, system architecture and deep multi-step reasoning powered by Claude Opus 5 & frontier intelligence on High.
Nxelle Codex
Medium · High
Coding only: write, debug, review and refactor code, plus shaders, powered by GLM 5.3 Sovereign & deep coding models.
Limits reset daily at local midnight. Frontier models (Claude Opus, GPT, Gemini Pro and similar) are capped at 100 messages/day on High so prices stay sustainable; Flash engines are not capped separately. After payment you receive an activation code — enter it under Sanctum Plan in the workspace. Annual billing saves 20%.
WHAT STAYS FREE — FOREVER
The Open-Core Sovereign Pledge
Not ready yet? Install the free Core and experience sovereign agent orchestration today →