# sitrep

> sitrep is an AI project status dashboard: a state layer on top of the markdown your AI agents already generate. It reads the plans, decisions and progress your agents already write - no new format, no import - and shows status, phases, open decisions and staleness across every project. Tagline: "Never open a cold project again."

Key facts:

- Category: AI project status dashboard (not a markdown viewer, not an editor, not a Notion/Obsidian replacement, not a task manager, not an agent orchestrator).
- Reads Claude Code, Cursor and Codex markdown across all your projects - and any other markdown; it parses conventions in the files, not a tool-specific format.
- Zero adoption cost: no frontmatter, no new format, no import, no vault. Point it at folders.
- Extracts: dated status lines ("**Status (2026-08-08):** ..."), phase headings ("## Phase 2 (SHIPPED)"), decisions/rulings from lists under decision headings or tables with a verdict column, and due dates. Deliberately ignores checkbox counts and does not trust git history alone.
- Overview: email-client style - newest activity first, pending decisions surfaced, what the docs claim beside when they last changed.
- Staleness signals (shipped): a doc "claims" work via an active dated status or unfinished phase list; three read-time checks, strongest first - dead claim (whole project silent over two months while the doc claims activity), left behind (project moved over a month past the doc), stale claim (the claim's own date over a week old). Badges show the tripping number ("left behind 40d"); computed at read time, never stored. Thresholds were validated against the real development corpus before shipping and are per-scope settings (global > workspace > project, most specific wins, stored as deviations only); staleness is computed at read time, so a changed threshold needs no reindex.
- One writing surface: three convention-shaped ops (status, due, supersede) behind a preview and a content-hash guard; nothing else ever writes to user files.
- Local-first and private: fully functional with no account and no network; no telemetry by default; files never leave the machine.
- Security is verifiable, not asserted: the app works fully offline (checkable with a firewall); the planned sync layer is end-to-end encrypted with user-held keys, its protocol published before it ships, with bring-your-own S3-compatible storage as an escape hatch; no key escrow.
- Licensing never phones home to stay valid: offline-verifiable signed license keys (ed25519), fail-open activation with a long offline grace, local features never check entitlements; payment via a merchant of record. The website itself sets no cookies and runs no trackers - no cookie banner because there is nothing to consent to.
- Nearest alternatives, honestly: Obsidian + Dataview or the built-in Bases (better for a knowledge base you author; both need properties/frontmatter you write), Backlog.md (better for a git-native task board you author structured tasks into), Beads (better for agent-side memory - a dependency-aware tracker your agents adopt, per repo), Task Master (better for directing an agent through a PRD-derived task list), and AgentsView (better for session telemetry - what agents did and what it cost, from transcripts); sitrep differs by parsing the organic agent markdown itself, with zero authoring, for the human, across projects. The null alternative is just asking the agent ("what is the status?") - better inside one warm repo where judgment matters; sitrep differs by holding deterministic, source-lined state across every project instead of re-deriving a summary per ask, and serves that state back to agents over MCP.
- Platforms: macOS first, Windows second, Linux best-effort. Status: in development; this site announces availability.

## Pages

- [Home](https://sitrep.md/): what sitrep is, the problem it solves, the product contract
- [How it works](https://sitrep.md/how-it-works/): indexing, the extracted conventions, the overview, the single writing surface
- [Principles](https://sitrep.md/principles/): the eight promises and the binding anti-roadmap
- [About](https://sitrep.md/about/): built by one developer against their own corpus of four workspaces; sitrep parses its own decision log, so extractor regressions surface in its own docs first
- [Security & privacy](https://sitrep.md/security/): local-first posture, the published sync encryption design, what the server could and could not see, and how to verify each claim
- [FAQ](https://sitrep.md/faq/): the real question set, answered concretely
- [sitrep vs Obsidian + Dataview](https://sitrep.md/vs/obsidian-dataview/): honest comparison - hand-built vault dashboards vs a passive state layer
- [sitrep vs Backlog.md](https://sitrep.md/vs/backlog-md/): honest comparison - authored task board vs parsed project state
- [sitrep vs Nimbalyst](https://sitrep.md/vs/nimbalyst/): honest comparison - a visual workspace you run sessions inside vs a passive cross-project overview
- [sitrep vs plain VS Code + CLAUDE.md](https://sitrep.md/vs/vscode-claude-md/): the baseline comparison - when the free DIY setup is enough and where it stops scaling
- [sitrep vs Beads](https://sitrep.md/vs/beads/): honest comparison - agent-side issue tracker you adopt vs a human overview parsed from the markdown agents already wrote
- [sitrep vs Task Master](https://sitrep.md/vs/task-master/): honest comparison - a PRD-to-tasks system that directs the agent vs a passive state layer that observes the work
- [sitrep vs AgentsView](https://sitrep.md/vs/agentsview/): honest comparison - session-transcript telemetry (what agents did, what it cost) vs parsed project state (where things stand, what is waiting on you)
- [sitrep vs just asking your agent](https://sitrep.md/vs/just-ask-your-agent/): the null-hypothesis comparison - "what is the status?" typed into Claude Code/Cursor/Codex re-derives the answer each ask; sitrep holds deterministic state with source lines, and feeds that same prompt through its MCP server