Command Palette

Search for a command to run...

UnylyUnyly
Весь каталог

Nauro

БесплатноНе проверен

The decision system for agentic engineering: keeps your project's decisions, rationale, and rejected paths in plain files and surfaces them to AI coding agents

GitHubEmbed

Описание

The decision system for agentic engineering: keeps your project's decisions, rationale, and rejected paths in plain files and surfaces them to AI coding agents before they plan or change code.

README

Give your agents the context code leaves out.

Nauro keeps current state, open questions, and human-approved project judgment in one record, ready for every agent you connect.

That record combines project scope, current state, and open questions with human-approved project judgment, including intent, goals, decisions, rationale, tradeoffs, and rejected paths.

Project judgment is the human-ratified part of the record. Context is the slice Nauro gives an agent for the work in front of it. Before connected agents plan or change work, Nauro surfaces that context. New or revised judgment becomes project truth only after you approve it.

The record belongs to the project and is available from Claude, Cursor, Codex, Perplexity, and other MCP clients.

PyPI Python License

Status: Stable (1.x). Semantic versioning covers the CLI, local stdio MCP contract, on-disk store format, and curated nauro-core import API. Cloud sync and hosted MCP are versioned separately.

Pareto example

https://github.com/user-attachments/assets/9e6c475b-c584-470b-84c2-12f01b3a425a

A real Codex session in Pareto, a reproducible mock project. Nauro retrieves the existing concurrency cap before the agent recommends a top-tier override. After approval, the agent records the decision and changes the code. This is one controlled example; recurring value in other projects remains unproven.

Install

uv tool install nauro

uv fetches its own Python, so Python does not need to be pre-installed. Install uv with curl -LsSf https://astral.sh/uv/install.sh | sh on macOS or Linux, or follow the PowerShell instructions on Windows. If Python 3.10 or newer is already installed, pipx install nauro and pip install nauro also work.

The first interactive run asks once about anonymous telemetry and defaults to no. nauro telemetry status shows the setting, and NAURO_TELEMETRY=0 disables telemetry and the prompt.

Try it in 30 seconds

The Pennykeep demo needs no account, MCP wiring, or agent restart:

mkdir -p /tmp/nauro-demo && cd /tmp/nauro-demo
nauro init --demo
nauro check-decision "Store dollar amounts as decimal numbers"

Pennykeep is a fictional local-first budgeting project with thirteen decisions. Its top match is:

D001 "Amounts stored in integer cents, never floating point"

The stored decision says to use integer cents because binary floating point cannot represent values such as 0.10 exactly. The agent reads the decision body, including its rationale and rejected path, and judges its relevance.

check-decision uses deterministic BM25 keyword retrieval and ranks results by keyword overlap rather than semantic meaning. It never blocks a change. The MCP check_decision tool returns the same result directly to a connected agent before planning.

Use Nauro on a real repo

Run adoption from the root of an existing Git repository:

cd your-repo
nauro adopt

The command registers the project, writes .nauro/config.json, and wires local MCP access plus the adoption skill for Claude Code, Cursor, and Codex. Restart the agent, then invoke /nauro-adopt.

The skill reads project documentation for rationale and inspects the repository as supporting evidence. When the files show what the project does but not why, it asks you instead of inventing a decision.

Review .nauro/config.json and commit it with the repo. Setup variants and optional tooling are covered in the quickstart and agents and skills guide.

If the repo already has a hand-written AGENTS.md, adoption and setup leave it in place and warn. Running nauro sync is the one action that overwrites it; a # Manual section survives the rewrite.

How the loop works

  1. Nauro orients the agent with project scope, current state, open questions, and relevant prior judgment.
  2. You and the agent clarify missing intent, constraints, or tradeoffs.
  3. If the work needs new or revised judgment, the agent drafts it and waits for your explicit approval.
  4. The agent plans, recommends, or implements with that context in view.
  5. The agent explains how the context shaped the result. You can accept or correct it, grant an exception, or reopen and supersede prior judgment.
  6. The agent reports meaningful completed progress as current state, so later connected agents inherit the updated state and approved judgment.

check_decision surfaces related records before a technical choice. propose_decision commits only the draft you have approved.

No model decides what your project believes. Retrieval remains advisory, and agent output cannot silently change project judgment. Current state and open questions share the record, but they do not use the judgment approval contract. The local record is plain Markdown in a folder you own, and cloud sync is optional.

When it fits

If a small repo plus a reliable AGENTS.md or CLAUDE.md keeps agents oriented, Nauro may be more than you need.

Nauro is intended for longer-lived work where context can be lost between sessions or needs to move between tools and repositories. If agents reliably consult committed decision files, those files may be enough. Built-in memories remain tied to one product; Nauro keeps the project record independent of the connected client.

Current limits:

  • The project record is necessarily partial. Nauro can surface only what has been recorded. Retrieval does not establish that an agent understands the project or that a decision still applies unchanged.
  • Keyword retrieval can miss a relevant decision phrased in different language. An optional embeddings index adds synonym matching.
  • Each connected-agent check adds an MCP read and remains advisory. Nauro does not force an agent to honor a result.
  • Nauro is not a backlog, task orchestrator, code indexer, or replacement for Git history.
  • Hosted projects are single-owner today. Nauro does not provide team governance or concurrent multi-writer authority.

Local and cloud access

Local use needs no account or cloud service. The record stays as plain Markdown on your machine unless you enable cloud sync. The hosted connector at https://mcp.nauro.ai/mcp and cloud setup are documented in Connect your agent.

Codex remote users must set mcp_oauth_callback_port = 8765 at the top of ~/.codex/config.toml. Without it, Codex selects a random callback port that Auth0 cannot accept.

Inspect the record

nauro graph writes one self-contained HTML file with Graph, Lineage, Timeline, and Browse views. By default, it embeds decision titles, metadata, open-question summaries, and full decision bodies. With --no-include-bodies, it keeps the titles, metadata, and question summaries but omits full decision bodies.

nauro doctor reports structural defects without editing the store.

Connected agents can request L0 for a concise orientation, L1 for a bounded working set, or L2 for a full dump. On mature stores, L2 can reach hundreds of thousands of tokens, so agents should start with L0 and retrieve exact records as needed.

Documentation

Uninstall

From an adopted repo:

nauro adopt --remove

This reverses adoption for the current repo. The Markdown store stays on disk unless you explicitly pass --purge-store for the project's last repo. Remove the CLI with uv tool uninstall nauro.

Development

The monorepo contains the nauro CLI and the reusable nauro-core library. See CLAUDE.md for the architecture.

uv sync --all-packages --all-extras
uv run pytest packages/nauro-core/tests/ -x -q
uv run pytest packages/nauro/tests/ -x -q

Report bugs and request features in GitHub Issues.

Apache 2.0.

Named for Peter Naur's Programming as Theory Building (1985).

from github.com/Nauro-AI/nauro

Установить Nauro в Claude Desktop, Claude Code, Cursor

Рекомендуется · одна команда, все IDE
unyly install nauro

Ставит в Claude Desktop, Claude Code, Cursor и VS Code — сам разбирается с npx, uvx и сборкой из исходников.

Впервые? Поставь CLI: curl -fsSL https://unyly.org/install | sh

Или настроить вручную

Выполни в терминале:

claude mcp add nauro -- uvx nauro

FAQ

Nauro MCP бесплатный?

Да, Nauro MCP бесплатный — установка в пару кликов через Unyly без оплаты.

Нужен ли API-ключ для Nauro?

Нет, Nauro работает без API-ключей и переменных окружения.

Nauro — hosted или self-hosted?

Self-hosted: сервер запускается локально на твоей машине командой из раздела установки.

Как установить Nauro в Claude Desktop, Claude Code или Cursor?

Открой Nauro на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.

Похожие MCP

Compare Nauro with

Не уверен что выбрать?

Найди свой стек за 60 секунд

Автор?

Embed-бейдж для README

Похожее

Все в категории development