# Tachikoma

> Your AI agents deserve real infrastructure.

Everyone's building agents. Nobody's building the infrastructure they need — permissions, audit trails, shared context, human-in-the-loop. Tachikoma is a 6-layer framework that gives AI agents the same governance you'd give a new hire: roles, permissions, audit trails, and a place in the org chart. Open source. Self-hosted. Zero vendor lock-in.

## The Problem

61% of enterprises have data not AI-ready (Enterprise Knowledge, 2025). Only 1 in 5 has mature agent governance (CSA, 2025). EU AI Act fines reach up to 35M EUR or 7% of global revenue. Most teams duct-tape 4+ tools — and still can't audit what their agents did last Tuesday.

## 6 Layers

### 1. U&A2CUI SDK — Collaborative UI
Users and agents as equal participants in the same session. CRDT-backed sync, permission-scoped views, 3 transport modes. The UI treats a human click and an agent decision identically.

### 2. Progressive Hive Context Protocol
Distributed living graph model. Inject context, watch behavior emerge. Cast injects signals, Gazer observes. Switch, Dispatch, Route for flow control. Zero central orchestrator.

### 3. Law&States Machine
Agents join your org chart. Roles define functional perimeters. BorrowACL (Rust) handles auth. StatefulRecord fuses state machines + event journal. Every action leaves an immutable audit trail.

### 4. Enterprise Agentic Layer
Agent objects (multi-provider), RLM (Recursive Living Context with Analyze/Execute/Enrich cycles), WOT (Workflow of Thought — directed graphs with typed data contracts). WOT Debugger with breakpoints and history replay.

### 5. Dynamic MCP Mesh
Per-token ACL. 3 modes per capability: on, off, borrow (use a secret but never read it). Hot-reload in <5s. MCPRegistry with health checks. ComposedTool chains MCP calls into Hive workflows.

### 6. Omnirepo
Your enterprise as a living context graph. 28+ XML entity types. Galaxy/System/Space hierarchy. Agents get Unix accounts like employees. Secrets encrypted until execution. Everything in git, traced in ClickHouse.

## Enterprise Stats (Sourced)

- 61% of enterprises have data not AI-ready — Enterprise Knowledge, 2025
- 3.4x more effective governance with dedicated platforms — Gartner, 2025
- 35M EUR max EU AI Act fines (or 7% revenue) — EU AI Act, 2026
- 40% of enterprise apps with AI agents by 2026 — Gartner, 2025
- 1 in 5 companies has mature agent governance — CSA, 2025

## Links

- Docs: https://holichrys.github.io/tachikoma
- GitHub: https://github.com/HoliChrys/tachikoma
- Full context: /llms-full.txt