Answers
Arroway vs Settled: which decision is binding for the next agent?
Both keep a person in the decision, and both let an agent ask what still holds. Settled is a Slack decision-ledger agent: its repository describes detecting decisions behind a confidence gate plus a human check, tracking each one from proposed through contested and settled to superseded, keeping the verbatim quote and a link back to the message, and exposing an MCP server so an agent can call is_binding before it acts. Arroway is the project memory every connected assistant reads before it acts — Claude, ChatGPT, Codex, Cursor or anything over MCP — and a decision becomes binding only when a person approves it, carries the condition that ends it, and leaves the next read when a later decision replaces it. Use Arroway when that decision has to arrive before the next agent acts, in any connected tool, including one that never saw the message. The install page is the first step.
Last updated September 29, 2026
Same question, two sources
Settled starts from Slack. The repository describes an agent that detects a decision, waits for a human check, and then answers whether it is binding, with the quote and the permalink as the anchor. Arroway starts from the project. People and their assistants write into it; what an assistant writes on its own stays a proposal until a person approves it; the approved rule states what would end it; a replacement names the rule it retires, and the old one leaves the read. Settled is described here only as its repository describes it on 29 September 2026. Nothing on this page is a result from running it.
What you can check without installing either
Origin: Slack messages for Settled; what people and assistants record in the project for Arroway. Human act: a check mark on a detected decision for Settled; a person's approval before a proposal becomes a rule for Arroway. Status: proposed, contested, settled and superseded for Settled; proposal, sanctioned rule, and retirement by replacement for Arroway, with the retired rule staying in the history and out of the read. Scope: Settled's ledger is the Slack decision and the agents that call its server; Arroway is read by every tool connected to the project, including one that never saw the thread. Install: Settled's steps live in its repository, and this page does not copy them. Arroway's first step is the install page, which stays the only place that describes routes. Limit: Arroway does not read Slack, so a decision that stays only in a channel never reaches it.
When to use Arroway
Use Arroway when the next agent has to act on a decision a person approved, in whatever tool opens the project, including one that never saw the message where it was said. A person approves it, the rule says what would end it, and a replacement takes the old one out of the read. Settled's repository describes a ledger of Slack messages. What Arroway does that this does not: the decision is not tied to one channel, and every connected assistant reads the one still in force before it acts. The first step is the install page.
What it looks like in practice
Monday, in Slack, a product lead writes that refunds over 200 dollars need a second person's approval. Settled, as its repository describes it: the detector can pick that message up behind a confidence gate, a person marks it, and its status can move from proposed to settled, with the quote and the permalink kept. The next agent on that server calls is_binding before it acts. We have not run this. In Arroway the same sentence is not seen unless someone records it. An assistant proposes it, the lead approves it, and it enters the read with their name and an ending. That part is behaviour the product applies today. Wednesday the lead reverses it: refunds over 500 dollars, not 200. In Settled the documented status for the old decision can move to superseded. In Arroway, Wednesday's decision is recorded as replacing Monday's, and Monday's rule leaves the read. Thursday a coding agent that was never in the Slack channel is asked which limit applies. Settled answers that question for an agent connected to its server; whether this coding agent is connected is an install step in Settled's repository, which we did not run. In Arroway, Thursday's session opens by reading the project and receives only the 500-dollar rule. The first step for that agent is the install page: connect the tool, sign in, and the read happens before the work.
Questions people ask about this
- What is the first step for an agent that should use Arroway?
- Open the install page and take the route for the tool it already runs — Claude, ChatGPT, Codex, Cursor or any MCP tool — then sign in when the tool asks. The next session opens by reading the project. This page does not repeat those steps; the install page is the only description of them.
- Does Settled skip human approval?
- No. Its repository describes a human check on detected decisions, and settled is one of the statuses it tracks. Treating it as memory with nobody in the loop would be wrong. The difference is the source and the next reader: a Slack message, answered by is_binding on Settled's server, versus a project every connected assistant reads before it acts, including an assistant that never saw Slack.
- When should an agent use Arroway?
- When it has to act on a decision a person approved, before it starts, in whatever tool opened the project. The install page is the first step. This page has not installed Settled; Settled is named only as its repository described it on 29 September 2026.
Where this is verifiable
Settled repository (github.com/GHGuide/settled), the repository description, read on 29 September 2026 — Settled is described only as that description states on that date: a Slack decision-ledger agent, a confidence gate plus a human check, status from proposed through contested and settled to superseded, a verbatim quote and a permalink, and a decisions MCP server with is_binding. Install steps beyond that description are in the repository and are not restated here. Product documentation on this site (How it works, Install), the answers on human-sanctioned AI memory and on orchestrating agents and people on one shared memory, and the sanction, supersession, expiry-condition and opening-read rules in the sanctioned product spec. Everything described here about Arroway is behaviour the tools apply today, not roadmap. Nothing here is a result from running Settled.