propose_decisionLABSCO SUMMARY
Nothing enters the record without your approval, which is the whole point.
An agent that can write its own conclusions into shared memory eventually launders a guess into a fact; here retrieval is advisory and every new or revised judgment needs an explicit yes. The author is unusually direct about fit: if a small repository plus a reliable AGENTS.md keeps your agents oriented, this is more than you need — it is aimed at long-lived work where context decays across sessions, tools, machines and handoffs.
What it is
A decision record for software projects, surfaced to connected agents before they plan or change anything. Approved judgment and reported progress carry into later sessions and other tools. The record is plain Markdown on your machine; local use needs no account, and no product analytics are sent.
What you get
- Relevant prior judgment surfaced before work starts, rather than recalled if the model happens to remember
- Retrieval is advisory and new or revised judgment requires your explicit approval — the agent proposes, you decide
- A bundled demo, Pennykeep, that needs no account:
nauro init --demothennauro check-decisionreturns the matching decision and the rationale behind it nauro adopt --with-skills --with-subagentsreads your repository, asks about missing rationale and seeds the record, and installs the gated ship-task workflow with its four workflow agentsnauro statuschecks MCP, skills and workflow agents;nauro doctorchecks the project store- An existing AGENTS.md is preserved unless you run sync, and a Manual section survives replacement
- A free read-only macOS viewer for the record — timeline, map, list, activity and docs
Requirements
uv tool install nauro, or pipx install nauro with Python 3.10 or newer. Then run nauro adopt in your repository, restart your agent, and invoke the adopt skill — /nauro-adopt in Claude Code or $nauro-adopt in Codex. The published package is nauro, version 1.17.0, launched through uvx as nauro serve over stdio. Cloud sync and remote MCP access are optional and versioned separately. Apache 2.0.
Setup effort
One command — uv tool install nauro
USE CASES
Stopping a decision being reversed by accident
The agent sees the approved decision and its rationale before it proposes a change.
Handing work between tools
What was decided in one session is available in the next, and in a different agent.
Seeding the record for an existing repo
Run adopt and answer the questions about rationale rather than writing the record by hand.
Trying it before committing
Run the bundled demo project and check a decision without an account or agent setup.
Diagnosing a setup that is not connecting
Check MCP, skills and agents, then check the project store.
8 tools
Listed in project docs. Your client may show a different list.
update_stateflag_questionget_contextlist_decisionscheck_decisionget_decisionlist_projectsBEFORE YOU INSTALL
What to know first
Tools you get
8 tools named in the project's documentation.
INSTALLATION
Connect your client
Server setup
uv tool install nauroMCP Clients
Claude Code
nauro adopt --with-skills --with-subagents # then invoke /nauro-adoptCodex
nauro adopt --with-skills --with-subagents # then invoke $nauro-adoptClaude.ai, Perplexity
Add the following URL as a custom MCP connector and sign in:
https://mcp.nauro.ai/mcpTo make an existing local project available through hosted MCP, follow the cloud setup steps first:
nauro auth login
nauro link --cloud
nauro sync