Blog — ARCANA

Deep DiveAug 19, 2026

Voice Input in Arcana: Fully Local Speech-to-Text

How Arcana ships a complete voice-to-text pipeline that never leaves your machine: whisper.cpp for transcription, Ollama for cleanup, and zero cloud dependencies.

EssayAug 23, 2026

Governed Autonomy: Proof Over Logs

Agents act. Someone must be able to answer for the acting. The case for governed autonomy, invariants over modes, delegation without escalation, and why evidence must outlive the session.

Read the manifesto

TutorialAug 10, 2026

Writing Your First Arcana Skill

Skills are reusable instruction sets that load on demand. Here is how to write one, from anatomy to distribution.

Deep DiveAug 5, 2026

Browser Automation in Arcana: MCP-Powered Chrome Control

How Arcana agents drive a real Chrome browser via the agent-browser MCP server: navigation, screenshots, form filling, and JavaScript execution.

ArchitectureJul 30, 2026

Subagents: How Arcana Delegates Work and Tracks Progress

The subagent system lets Arcana spawn child sessions with live progress, approval attribution, and self-healing resolution when things go wrong.

CustomizationJul 25, 2026

Theming Arcana: 10 Built-In Themes and Custom Token Configuration

Arcana ships with 10 TUI themes and supports custom themes with ~75 tokens and automatic contrast floor enforcement.

ReleaseAug 19, 2026

v0.4.0: Governance Overhaul, Voice Input, and a Scriptable CLI

The permission system is rebuilt around configurable policy, the TUI gains fully local speech-to-text, and every CLI command speaks JSON with deterministic exit codes. Full ledger inside.

Read the notes

InfrastructureJul 20, 2026

The Proxy API: Your Keys, Your Rules

Arcana proxies API requests through a local relay so you can use your own provider keys with the free tier and control exactly where requests go.

TutorialJul 15, 2026

Scriptable CLI: JSON Output and Deterministic Exit Codes

Every Arcana CLI command supports --json output and returns deterministic exit codes, making it safe to use in scripts, CI pipelines, and automation.

Deep DiveJul 10, 2026

Session Compaction: Keeping Long Conversations Alive

When conversations get long, Arcana automatically compacts context to stay within model limits without losing the important parts.

SafetyJul 5, 2026

File Edit Guard: Safety Rails for Automated Code Changes

Arcana's file-edit guard system adds confirmation prompts before file modifications, with visual diffs and rollback support.

EvidenceJul 26, 2026

Arcana vs OpenCode vs Hermes: The Multi-Agent CLI Benchmark

Same model, same tasks, single attempt, cold start. Nine metrics across coding and domain suites, fastest on every task at 28.4s median, with full reproducibility via SHA256-locked fixtures.

See the results

GuideJun 28, 2026

The Arcana Free Tier: What You Get and How Quotas Work

Arcana offers a genuine free tier with session management, quota tracking, and model discovery. Here is how it works under the hood.

ArchitectureJun 22, 2026

MCP and Tools: How Arcana Extends Its Capabilities

The Model Context Protocol lets Arcana agents use any tool that speaks MCP. Here is how built-in tools, MCP servers, and custom tools work together.

SecurityJun 18, 2026

Trust Boundaries: Arcana's Security Model Explained

Arcana draws explicit trust boundaries between the agent, the workspace, the network, and external services. Here is how they work and why they matter.

Deep DiveJun 15, 2026

Memory: How Arcana Remembers Across Sessions

Arcana's memory system persists context across sessions using local storage, letting agents pick up where you left off without re-explaining everything.

ArchitectureJun 10, 2026

Desktop Integration: The Governance Bridge Between TUI and GUI

Arcana's TUI delegates permission prompts to a desktop GUI when available, creating a governance bridge between terminal-native work and visual approval.

TutorialJun 5, 2026