octave-mcp 0.3.0

OCTAVE MCP Server - Lenient-to-Canonical OCTAVE pipeline with schema validation

OCTAVE MCP Server

Production-grade MCP server implementing the OCTAVE document protocol for deterministic, auditable AI communication.

Table of Contents

• For AI Agents
• What It Does
• Quick Start
• Installation
• MCP Tools
• When OCTAVE Helps
• Documentation
• Contributing

---

For AI Agents

===AGENT_BOOTSTRAP===
GUIDANCE::AGENTS.oct.md
QUALITY_GATES::[mypy,ruff,black,pytest]
DEV_SETUP::docs/guides/development-setup.md
SPECS::specs/README.oct.md
IMMUTABLES::[I1,I2,I3,I4,I5]
===END===

---

What It Does

This repository ships the OCTAVE MCP Server—a Model Context Protocol implementation that exposes the OCTAVE document protocol as deterministic tools. The MCP layer is delivery plumbing; the value is the OCTAVE protocol itself.

OCTAVE (Olympian Common Text And Vocabulary Engine) is a deterministic document format and control plane for LLM systems. It keeps meaning durable when text is compressed, routed between agents, or projected into different views.

• Non-reasoning: OCTAVE never guesses intent; it only accepts, validates, and projects what you declare.
• Lenient → canonical: ASCII aliases and flexible whitespace are accepted on ingest, then normalized to canonical Unicode with a logged repair tier.
• Schema-anchored: Data and schema blocks travel together so routing and validation remain explicit.
• Auditable loss: Compression and projections must declare what was dropped; nothing is silently tightened or weakened.

Language, operators, and readability

• Syntax: Unicode-first operators (→, ⊕, ⧺, ⇌, ∨, ∧, §) with ASCII aliases (->, +, ~, vs, |, &, §) keep documents compact while staying typable everywhere.
• Vocabulary: Mythological terms are deliberate compression shorthands (e.g., ICARIAN, SISYPHEAN, HUBRIS→NEMESIS) that pack multiple related concepts into single tokens for higher semantic density.
• Authoring: Humans typically write in the lenient view and rely on octave validate to normalize into canonical Unicode; both views stay human-auditable.

See the protocol specs in specs/ for the precise operators, envelopes, and schema rules (v5.1.0).

What this server provides

octave-mcp bundles the OCTAVE tooling as MCP tools and a CLI.

• 3 MCP tools: octave_validate, octave_write, octave_eject
• CLI commands: octave validate, octave write, octave eject
• Deterministic: Non-reasoning control plane for syntax, validation, projection
• Loss accounting: Every transformation logged with audit trails
• Schema-anchored: Validation travels with documents

When OCTAVE Helps

Use OCTAVE when documents must survive multiple agent/tool hops, repeated compression, or auditing:

• Coordination briefs, decision logs, policy artifacts that circulate between agents
• Reusable prompts or RAG artifacts needing stable structure across context windows
• Documents mixing prose with routing/targets (e.g., §targets for tools or indexes)

Proven efficiency:

• 54–68% token reduction vs equivalent JSON while preserving fidelity
• 90.7% comprehension rate (zero-shot across Claude, GPT-4o, Gemini)
• Higher quality outputs (9.3/10 vs 8.3/10) when compression is explicit

See docs/research/ for benchmarks and validation studies.

Installation

PyPI:

pip install octave-mcp
# or
uv pip install octave-mcp

From source:

git clone https://github.com/elevanaltd/octave-mcp.git
cd octave-mcp
uv pip install -e ".[dev]"

Quick Start

CLI

# Validate and normalize to canonical form
octave validate document.oct.md --schema DECISION_LOG

# Write with validation (from content)
echo "===DOC===\nKEY::value" | octave write output.oct.md --stdin --schema META

# Project to a view/format
octave eject document.oct.md --mode executive --format markdown

MCP Setup

Add to Claude Desktop (claude_desktop_config.json) or Claude Code (~/.claude.json):

{
  "mcpServers": {
    "octave": {
      "command": "octave-mcp-server"
    }
  }
}

Or use the setup script:

./setup-mcp.sh --all              # Configure all clients
./setup-mcp.sh --show-config      # Show config for copy/paste

See docs/mcp-configuration.md for advanced configuration.

MCP Tools

Tool	Purpose
octave_validate	Schema validation + canonical normalization
octave_write	Unified file creation/modification (content OR delta changes)
octave_eject	Format projection (octave, json, yaml, markdown)

Repair Tiers

OCTAVE separates semantic decisions from mechanical guarantees through three repair tiers:

• NORMALIZATION (always): ASCII→Unicode, whitespace (semantics preserved)
• REPAIR (opt-in): Schema-bounded coercions (enum casefold, type conversion)
• FORBIDDEN (never): Semantic inference, structure invention, target guessing

Every transformation is logged for auditability.

Documentation

Doc	Content
Usage Guide	CLI, MCP, and API examples
API Reference	Python API documentation
MCP Configuration	Client setup and integration
Protocol Specs	Canonical operators, envelopes, schema rules (v5.1.0)
Development Setup	Dev environment, testing, quality gates
Architecture	Decision records and design docs
Research	Benchmarks and validation studies

Architecture Immutables

ID	Principle
I1	Syntactic Fidelity — normalization alters syntax, never semantics
I2	Deterministic Absence — distinguish absent vs null vs default
I3	Mirror Constraint — reflect only what's present, create nothing
I4	Transform Auditability — log every transformation with stable IDs
I5	Schema Sovereignty — validation status visible in output

Contributing

See CONTRIBUTING.md for development setup, testing, and guidelines.

# Quick dev setup
git clone https://github.com/elevanaltd/octave-mcp.git
cd octave-mcp
uv venv && source .venv/bin/activate
uv pip install -e ".[dev]"

# Run tests (706 passing, 87% coverage)
pytest

# Quality checks
ruff check src tests && mypy src && black --check src tests

License

Apache-2.0 — Built with MCP Python SDK.