Open-source · BSL · EU AI Act Art. 12 ready

The Company Brain
your humans and agents share.

Deterministic impact analysis. Immutable audit FK. Self-host. Agent-native. BSL classic license — auto-converts Apache 2.0 in 4 years.

16
KG edge types
FK
Immutable audit
L1→L5
Brain layers
BSL
EU ready
Why MarvisX

Three problems your team hits every day.

MarvisX answers each with a deterministic primitive — not a probabilistic guess.

🜨

Cross-project impact analysis

"If I deprecate API v1, what breaks across 23 projects?"

Two days of grep — or two seconds of graph_impact.

Deterministic 16-edge knowledge graph. No LLM guessing — just edges. Functions, modules, tasks, PRs, decisions, learnings, all linked. Depth-2, every downstream caller, milliseconds.

Capability M3.2 · UNIQUE in 25-player landscape
🜔

Decision audit chain

"Who approved this deploy on March 14? Due diligence asks tomorrow."

audit_log FK-linked task / PR / user, append-only. Two seconds of query, not two weeks of Slack and Notion archaeology.

EU AI Act Art. 12 ready. DORA-aligned. Series-B due-diligence ready.

Capability M2.4 + M7.1 · UNIQUE in 25-player landscape
🜂

Agent-native knowledge

"My agent forgot context from yesterday's session."

MCP-native protocol. Compatible with Claude Code, Codex, OpenCode. Constitution hooks enforce safety — no agent can hotfix prod, bypass project_id visibility, or skip task tracking.

Your agent and you share the same Brain. Zero re-onboard cost.

Domain M6 + Constitution 5 immutable rules
How it works

8 domains. 43 capabilities. One Brain.

MarvisX is structured as 8 functional domains — each solving one class of problem, each individually deployable.

M1 CAPTURE

Ingest substrate

Files, meetings, sessions, connectors. 4 parser lanes: OCR / docparse / transcribe / vision.

M2 STORE

Substrate of truth

16-edge knowledge graph + canonical doc store + immutable audit FK. The layer every query lands on.

M3 RETRIEVE

Find anything

Semantic (Voyage) + deterministic graph traversal + time-travel as_of + PR impact.

M4 REFLECT

Brain loop

L1 Substrate → L2 Digest+Journal → L3 Drift checker → L4 Memory ops → L5 Learn findings.

M5 EXECUTE

Tasks · PRs · Sessions

Triage, multi-agent sessions, decision propagation, in-house workflow orchestrator.

M6 AGENT-NATIVE

Agents first-class

MCP server + Constitution 5 bash hooks + quality gates + agent registry.

M7 TRUST

Audit-grade

RBAC 4 roles · Compliance view · Safety preflight · BYOK.

M8 PRODUCTIZATION

Ship layer

Self-host wizard · Multi-tenant · Billing · Site · Versioning train.

Pricing

One Brain. Three deployment paths.

From €0 self-host to whitelabel enterprise engagement.

marvisx-OSS
€0
forever · BSL classic

Self-host. Full feature parity. For developers, scaleups, agent-native builders.

  • 16-edge knowledge graph cross-project
  • FK-linked immutable audit log
  • Brain reflection L1→L5
  • Constitution 5 immutable hooks
  • MCP server (Claude / Codex / OpenCode)
  • Time-travel as_of queries
  • Self-host on your hardware
Auto-converts to Apache 2.0 after 4 years.
marvisx-Agent
from €25
/ seat / month · waitlist

Cloud SaaS. Chat-first. Mass market. For executives, PMs, partnership teams.

  • Hosted multi-tenant Cloud
  • chat-first UI (chat.* subdomain)
  • Per-seat billing
  • BYOK opt-in (your keys, or pool managed)
  • Hosted MCP for your agents
  • Pool-managed LLM default
Pre-launch waitlist. Final pricing post-OSS traction.
marvisx-Enterprise
Custom
from €50K / engagement

On-prem · Air-gap · Whitelabel · EU compliance custom. For legal, healthcare, banking, fintech.

  • Whitelabel branding per engagement
  • Air-gap deploy if required
  • EU AI Act Art. 12 custom retrofit
  • GDPR per-tenant compliance
  • Dedicated support engineer
  • Custom connectors (Slack / Teams / SharePoint / Linear)
  • Cryptographic audit chain
Active whitelabel engagements available on request.
▾ Compare all features by tier
Capability OSS Agent Enterprise
16-edge knowledge graph
Audit FK immutable● + crypto chain
Brain 5-layer reflection◐ + custom patterns
MCP serveropt
Constitution hooksN/A
Multi-tenant isolationN/A
Chat UIN/Aopt
WhitelabelN/AN/A
Air-gap deployN/A
EU AI Act Art. 12● custom retrofit

● FULL · ◐ PARTIAL · ◇ VISION · N/A not applicable

See MarvisX in action

90 seconds. Three queries. Real graph.

Video walkthrough coming with v1 launch
  • Q3 impact — "If I refactor this function, what breaks?"
  • Q6 audit — "Who approved deploy on date X?"
  • Brain reflection — drift signal → finding → human approve
About

Built to solve a real problem.

MarvisX is built by Emilio Martucci.

The project started in early 2026 to solve a problem I kept hitting: AI agents lose context across sessions, humans lose context across projects, and nobody can answer simple audit questions like "who approved this on March 14?"

MarvisX is a Company Brain: one queryable substrate for humans and agents to share. Open-source under BSL classic, self-hostable, EU compliance-ready.

The project is part of the ZNEXT Founders Program (founder-led, pre-seed).

— Emilio Martucci