An agent that notices when a decision happens in Slack, logs it with a permalink, and answers “what did we decide about X?” — with receipts. No commands, no forms, no /log-decision rituals. People just talk.
Receipts, not vibes. Every answer links to the thread where the decision actually happened — you never take the agent's word for it.
Three weeks later someone asks “wait, didn't we already pick a vendor for this?” — cue twenty minutes of scroll-archaeology, or worse, re-litigating a settled question.
Decision-log tools exist — every one of them relies on a human remembering to write things down. Precedent notices on its own.
Nobody runs a command. The agent listens, extracts, links, and follows up — the team just talks.
Precedent hears every message and says nothing. People talk in bursts — so it waits until a thread has been quiet for 90 seconds, reads the whole thing, and asks Claude one question: did a decision happen here?
Confidence ≥ 0.7 → saved with a permalink, a 📌 lands on the thread, one line confirms it. A debate, filler chatter, a “let's revisit after the interviews” → nothing. Deferrals aren't decisions.
Open the assistant panel — or @mention it anywhere. Two lookups run at once: the structured decision log, and live workspace search through Slack's Real-Time Search API, semantic first with keyword fallback.
Claude folds both into ≤120 words under one hard rule: answer only from evidence — never invent. When nothing was ever decided, it says so and points to the closest discussion instead. Try the third question.
The extractor sees the existing log, so when a new decision replaces an old one they're linked automatically — nobody files anything. Ask about email and you get the current answer and the story: what it replaced, and why.
Superseded decisions are struck through, never deleted.
Every action item is tracked. On weekday mornings, owners get a DM about what's still open — the buttons update the message in place. And the whole log is an MCP server, so Claude Desktop — or any agent — can ask your team's memory directly.
Built for the Slack Agent Builder Challenge — New Slack Agent track.