🧠 AI Development Intelligence

Ship faster.
Measure everything.

Sensei captures how your team uses AI — context quality, session costs, first-try-right rates — and feeds that intelligence back into your workflow.

Get started Usage guide →
62%
avg FTR improvement
$0.18
median task cost
4.2×
faster with context
38%
token savings via cache
Setup

Install Sensei

Download the desktop app — it installs the daemon, CLI, and MCP server on first run.

Download for macOS Apple Silicon

Sensei_aarch64.dmg

Download
🍺 Homebrew macOS / Linux
brew install --cask mizukisu/tap/sensei-app

Or CLI only: brew install mizukisu/tap/sensei

⚡ Included
  • ✓ Desktop app with setup wizard
  • ✓ senseid indexer daemon
  • ✓ sensei CLI
  • ✓ sensei-mcp server
📦 All downloads
macOS Apple Silicon detected macOS Intel Linux x86_64 Linux ARM64 Windows x86_64
What sensei does

The observability layer for AI-assisted development

Like Code Climate for code quality — sensei gives you the metrics to understand and improve how your team uses AI.

🎯

First-Try-Right Scoring

Track how often AI completes tasks on the first attempt. Per-repo, per-agent, per-team.

💰

Cost Intelligence

Token-level cost per task, session, and repo. See exactly where money is spent.

📦

Context Packing

Smart token-budgeted context using BM25 + semantic search. Right code, right size.

🔄

Session Continuity

Automatic crash recovery and progress snapshots. Never lose context mid-task.

📚

Doc Drift Detection

Traceability matrix linking code to docs. Flags stale documentation automatically.

🧩

Plugin Ecosystem

Skills, commands, and hooks that inject guardrails into every AI session.

How it works

Architecture

AI platforms connect via MCP. Desktop and CLI talk directly over HTTP. All data stays local.

senseid

Indexer daemon — watches repos, builds code graph, serves HTTP API

sensei-mcp

MCP adapter — translates AI tool calls into HTTP

sensei CLI

Manages installation, config, scanning

Desktop

Project navigator, graph viewer, indexer control

Reference

Usage Guide

CLI commands, daemon management, and MCP tools.

CLI Commands

🔧 Setup
sensei configure Detect AI coding platforms and generate config
sensei install Install hooks, skills, and MCP server
sensei install --acp claude-code Install for a specific platform
sensei uninstall Remove sensei from all platforms
🔍 Indexing
sensei scan ~/Developer Scan a folder and index all repos
sensei add-lib react Add library docs (auto-discovers llms.txt)

Daemon Management

senseid start Start as background daemon
senseid stop Stop the daemon
senseid status Show version and PID
senseid logs Tail the last 50 lines

MCP Tools

Available to your AI coding platform when connected via sensei-mcp.

🎯
get_session_context

Resume from last checkpoint, load repo orientation

📦
context_pack

Load symbols relevant to a query — token-budgeted

🔎
search

Semantic search across the entire codebase

📄
load_context

Load a specific file with symbol extraction

💾
take_snapshot

Record progress for interruption recovery

✅
checkpoint

Close session with outcome and FTR scoring

⬡ sensei · Setup · Reference · GitHub