Decision Log
인간과 AI 에이전트를 위한 추가 전용 결정 기록 시스템으로, 출처, 대체, 검색 및 감사된 MCP 작업을 제공합니다.
문서
Decision Log
Make every decision retrievable.
Decision Log is the durable system of record for decisions made by people and AI agents. It keeps the decision, rationale, Actor, source provenance, supersession history, and current status together so teams and tools can recover what was decided and why.
What Decision Log records
Decision meaning and provenance
- Immutable Decisions with stable keys, rationale, and Actor attribution.
- DecisionSource and SourceContent provenance for supporting evidence.
Change and review
- Explicit DecisionSupersession relationships instead of silent meaning edits.
- Reviewable DecisionCandidates for AI extraction and ingestion pipelines.
Auditability
- Append-only AuditEvents for meaningful actions and authorization decisions.
Decision Log stays deliberately narrow. It is not a project manager, ticket tracker, wiki, chat product, meeting transcription service, or generic knowledge base.
One record, four first-class surfaces
Human surface
- Web application for people who create, review, supersede, search, and audit decisions.
Agent and developer surfaces
- GraphQL API as the canonical programmatic schema.
- OpenAPI REST facade, generated from and delegated to the GraphQL schema.
- Authenticated MCP endpoint for permission-scoped agent tools and resources.
- Public developer repository for the official SDK, CLI, Agent Skills, and Agent Plugin.
The public CLI and SDK use the same REST façade, GraphQL service layer, and authorization model.
Safe agent workflow
- Retrieve the current Decision before acting.
- Inspect provenance and the supersession chain when the reason for change matters.
- Submit extracted or uncertain meaning as a DecisionCandidate for review.
- Create a new Decision and explicit supersession when approved meaning changes.
- Preserve Organization and Workspace scope, least privilege, Actor attribution, and audit context.