> ## Documentation Index
> Fetch the complete documentation index at: https://cofounder.appeeky.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Decision Log

> Browse the full history of what the agent recommended, what you did, and what happened.

The Decision Log is a chronological record of every action the agent has proposed, joined with the outcome (when measured).

## Why it exists

Without a decision log, the agent (and you) have no idea whether suggestions actually worked. With one, you can:

* Hold the agent accountable: *"You said this would lift downloads — did it?"*
* Spot patterns: *"Every metadata change in Q1 helped; nothing in Q3 did."*
* Onboard new teammates with the full reasoning history.
* Rebuild trust gradually by promoting auto-approve only on action types with consistent good outcomes.

## Where it lives

A SQL view, `cofounder_decision_log`, joins:

* `cofounder_actions` — the proposal: type, payload, status, timestamps
* `cofounder_memory_episodes` — the outcome: summary, before/after metrics

## What you see

Surfaced inline in the briefing under *"📜 Decision Log"*:

```
## 📜 Decision Log (last 12 decisions)

Applied (4):
  • [asc_metadata_update] Update subtitle (en-US) — applied 18d ago → downloads +63% | ASO 38 → 41
  • [asc_review_reply] Reply to review: "Crashes" — applied 14d ago → outcome pending
  • [linear_issue] Investigate v3.2.1 crash — applied 12d ago → outcome pending
  • [asc_metadata_update] Update keywords (en-US) — applied 6d ago → outcome pending

Still pending (3):
  • [asc_metadata_update] Update subtitle (de-DE) — proposed 4d ago
  • [linear_issue] Add localization for MX — proposed 3d ago
  • [github_issue] Fix screenshot generator — proposed 2d ago
```

## API

```bash theme={"theme":{"light":"github-light","dark":"github-dark"}}
GET /v1/cofounder/decision-log?appId=6479581124&lookbackDays=60
```

Response:

```jsonc theme={"theme":{"light":"github-light","dark":"github-dark"}}
{
  "data": {
    "decisions": [
      {
        "actionId": "uuid",
        "actionType": "asc_metadata_update",
        "title": "Update subtitle (en-US)",
        "status": "applied",
        "proposedAt": "2026-04-12T08:00:00Z",
        "appliedAt": "2026-04-12T08:42:00Z",
        "outcomeSummary": "Subtitle change deployed; impressions +63% over 14 days.",
        "measuredOutcome": "downloads +63.0% | ASO 38 → 41"
      }
    ]
  }
}
```

## How the agent uses it

The decision log feeds back into the reasoner as memory context. When the agent considers proposing a similar action, it sees:

* Whether you've ignored the same kind of suggestion before
* What the measured outcome of past attempts was
* Whether the action type has been net-positive for your app

This is what makes the agent feel like it's *learning* over time rather than starting fresh every morning.