# Context Ledger v3.0 by Godswarm

> Epistemic engineering memory with decision history, superseded-rule blocking ([ACTIVE] / [WITHDRAWN] / [DISPUTED]), Day-1 CLI (`context-ledger init / audit / query / doctor / sync-rules` v3.1), and verifiable line citations (source_locator) for Claude Code, Cursor, Codex CLI, Antigravity, and Godswarm agents.

## Key Metrics (Verified ALM-009)
- Gate A Frozen Regressions (v2.1 + v2.2 + v2.3 across Claude Sonnet, Claude Fable, ChatGPT GPT-6-Astra): **99.64% (`275/276`, 0 errors)** curated and **100.0% (`144/144`, 0 manual ops)** zero-manual cold auto-ingest.
- 3,200-File Posting Index Latency (~6.3M est tokens / 25.21M chars): **`0.159 ms p95`** Fast-Path (145 injected tokens), **`8.952 ms p95`** Escalation-Path.
- Sealed Blind 6-Domain Holdout-2 (OOD): **`91.67%` (`99/108`)** vs Chunk-RAG **`63.89%`** (`+27.78 pp`) and File-Search Agent **`55.56%`** (`+36.11 pp`), reducing API input tokens by **`6.29x`** (`13,222` vs `83,131`).
- False-Premise Protocol Guard (T08 across 120k..10M): **`12/12 (100%)`** vs Chunk-RAG **`0/12 (0%)`**.
- Day-1 CLI v3.1 (`context-ledger init --ide all && context-ledger audit . --ingest && context-ledger sync-rules && context-ledger doctor --roi`, plus `--target AGENTS.md` for Cursor/Codex/Antigravity and `context-ledger query`): Zero-LLM repository audit finds CLAUDE.md/AGENTS.md/.cursorrules contradictions & superseded ADR references in <500 ms.

## Core Endpoints & Discovery
- [Product Landing (RU default / EN toggle)](https://ledger.godswarm.ai/): Full interactive simulator, ROI & payback calculator, IDE configs, and Early Access form.
- [Download Standalone Offline HTML](https://ledger.godswarm.ai/download/landing.html): Single-file self-contained browser deliverable.
- [Full Technical & Commercial Spec](https://ledger.godswarm.ai/llms-full.txt): Pricing tiers, MCP tools, Belnap-Dunn truth model, and integration examples.
- [MCP Discovery Manifest](https://ledger.godswarm.ai/.well-known/mcp.json): 7 MCP tools and 9 resources/templates.
- [A2A Agent Card](https://ledger.godswarm.ai/.well-known/agent.json): Machine-readable capabilities and pricing metadata.
- [Pricing & Benchmarks JSON](https://ledger.godswarm.ai/api/pricing) | [Benchmarks](https://ledger.godswarm.ai/api/benchmarks)
- [Lead / Early Access Intake](https://ledger.godswarm.ai/api/access-request): POST JSON (`name`, `email`, `org`, `role`, `plan`, `scale`, `use_case`) returning persistent `CLR-*` receipt.
- [Parent Ecosystem](https://godswarm.ai): Stigmergic parallel coding swarm powered natively by Context Ledger.

## Pricing (USD, -20% Annual, Closed Commercial v3.0 Core)
- **Starter / Trial ($0, 14 days):** Full Fast-Path + Escalation-Path, 1 user, 1 workspace up to 25M chars / 2,000 queries.
- **Pro Solo ($29/mo Early Access, $23.20/mo annual):** 1 user, 5 workspaces x 25M chars, Cold Auto-Ingest 100%, full bitemporal audit, up to 3 devices.
- **Team + Godswarm Bundle ($69/seat/mo, $55.20/seat/mo annual, min 3 seats):** Shared team [WITHDRAWN]/[DISPUTED] ledger, roles, audit log, CI hooks + included Godswarm Swarm license (standalone Team Ledger option $49/seat/mo).
- **Enterprise On-Prem / VPC (from $2,400/mo / $28,800/yr):** Air-gapped / VPC deployment under B2B contract, SSO/OIDC, SHA-256 compliance ledger, SLA 99.9%, optional $3,000 30-day Gate A pilot credited to Year 1.

## Architecture (Layers 0 to 6)
1. **Layer 0 (Immutable Evidence Log):** Append-only SHA-256 anchored segments (`source_locator` exact file/line pointers).
2. **Layer 1 (Awake Hot Path & Strategy 097 Rules):** Sub-millisecond compiled pattern rules (`0` LLM tokens) + compound entity-namespace isolation (`service-core` vs `service-core-eu` / `-gov`).
3. **Layer 2-3 (Bitemporal Belnap-Dunn Ledger):** Tracks `valid_from` / `valid_to` / `recorded_at` and 4-valued truth status (`TRUE`, `FALSE`, `BOTH`/`DISPUTED`, `NEITHER`/`NOT_ESTABLISHED`).
4. **Sleep-Time Compute (`sleep_consolidate`):** Passive `StructuralSpecExtractor` + targeted supersession linking lifting Cold Auto-Ingest to `100.0%` (`144/144`).
5. **Layer 4-5 (Adaptive `query_v3`):** `0.159 ms p95` Fast-Path (~145-1,350 tok) with deterministic risk escalation (`T1` attr gap, `T2` namespace collision, `T3` bitemporal delta, `T4` narrative MMR, `T5` false-premise check).
6. **Layer 6 (MCP + Hooks + Godswarm):** Works via local `stdio` and loopback `SSE` MCP (`BYO-compute`, zero code leaves client), plus built-in Godswarm swarm coordination.

## Request Early Access via API
```bash
curl -X POST https://ledger.godswarm.ai/api/access-request \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Alex Petrov",
    "email": "alex@acme.dev",
    "org": "Acme Engineering",
    "role": "Staff Engineer",
    "plan": "team_bundle",
    "billing": "annual",
    "scale": "1m_5m",
    "ide": "claude_code_godswarm",
    "use_case": "Eliminate stale architecture decisions across 12 engineers using Claude Code and Godswarm."
  }'
```
